http://www.redbooks.ibm.com/pubs/pdfs/redbooks/sg247040 (3/02

http://www.redbooks.ibm.com/pubs/pdfs/redbooks/sg247040 (3/02
IBM
Front cover
pSeries
IBM
670
70 and pSeries 690
System Handbook
Component-based description of the
hardware architecture
How to use e-config to configure
pSeries 670 and pSeries 690
Capacity Upgrade on
Demand explained
Keigo Matsubara
Alvin Hua Juay Teng
Bruno Blanchard
ibm.com/redbooks
International Technical Support Organization
IBM ~ pSeries 670 and pSeries 690
System Handbook
May 2003
SG24-7040-02
Note: Before using this information and the product it supports, read the information in
“Notices” on page xiii.
Third Edition (May 2003)
This edition applies to IBM ^ pSeries 670 Model 671 and IBM ^ pSeries 690 for
use with the AIX 5L Version 5.1 (program number 5765-E61) and AIX 5L Version 5.2 (program
number 5765-E62).
Note: This book is based on a pre-GA version of product IBM Hardware Management Console
for pSeries software Release 3.2 and may not apply when the product becomes generally
available. We recommend that you consult the product documentation or follow-on versions of
this redbook for more current information.
© Copyright International Business Machines Corporation 2002, 2003. All rights reserved.
Note to U.S. Government Users Restricted Rights -- Use, duplication or disclosure restricted by GSA ADP Schedule
Contract with IBM Corp.
Contents
Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Notices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii
Trademarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
The team that wrote this redbook. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
Become a published author . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvi
Comments welcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvii
Summary of changes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
April 2003, Third Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
October 2002, Second Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix
March 2002, First Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxi
Chapter 1. Overview of the pSeries 670 and pSeries 690 . . . . . . . . . . . . . . 1
1.1 What’s new in the pSeries 670 and pSeries 690. . . . . . . . . . . . . . . . . . . . . 2
1.2 pSeries 670 and pSeries 690 characteristics . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Microprocessor technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Memory subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.3 I/O drawer subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.4 Hardware technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Logical partitioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.4 Dynamic logical partitioning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.5 General overview of pSeries 670 and pSeries 690 . . . . . . . . . . . . . . . . . . 10
1.6 Market positioning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.7 Supported operating systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7.1 AIX 5L Version 5.1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.7.2 AIX 5L Version 5.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.7.3 Linux - SuSE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.7.4 Comparison of RAS supported features . . . . . . . . . . . . . . . . . . . . . . 14
1.7.5 Installation and backup of the operating systems . . . . . . . . . . . . . . . 15
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690 . . 17
2.1 What’s new in the pSeries 670 and pSeries 690. . . . . . . . . . . . . . . . . . . . 18
2.2 Modular design of the pSeries 670 and pSeries 690 . . . . . . . . . . . . . . . . 18
2.3 Central Electronics Complex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
© Copyright IBM Corp. 2003. All rights reserved.
iii
2.3.1 POWER4 processor and MCM packaging . . . . . . . . . . . . . . . . . . . . 22
2.3.2 Memory subsystem for pSeries 690 . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.3 MCMs and GX slots relationship for pSeries 690 . . . . . . . . . . . . . . . 39
2.3.4 Memory subsystem for pSeries 670 . . . . . . . . . . . . . . . . . . . . . . . . . 42
2.3.5 MCMs and GX slots relationship for pSeries 670 . . . . . . . . . . . . . . . 44
2.3.6 I/O books. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.3.7 Service processor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.4 I/O subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.4.1 I/O drawer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.4.2 I/O subsystem communication and monitoring . . . . . . . . . . . . . . . . . 56
2.4.3 I/O drawer physical placement order . . . . . . . . . . . . . . . . . . . . . . . . 65
2.4.4 Media drawer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2.5 Power subsystem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
2.5.1 Bulk power assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
2.5.2 Internal battery feature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
2.5.3 Cooling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
2.6 IBM Hardware Management Console for pSeries . . . . . . . . . . . . . . . . . . . 75
Chapter 3. Using the IBM Configurator for e-business . . . . . . . . . . . . . . . 79
3.1 What’s new with e-config. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.2 Configuration rules for pSeries 670 and pSeries 690 . . . . . . . . . . . . . . . . 81
3.2.1 Minimum configuration for the pSeries 670 and pSeries 690 . . . . . . 83
3.2.2 LPAR considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.2.3 Processor configuration rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.2.4 Memory configuration rules. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
3.2.5 I/O books. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
3.2.6 Media drawer configuration rules . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.2.7 I/O drawer configuration rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.2.8 I/O loops and cabling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.2.9 Graphics console configuration rules . . . . . . . . . . . . . . . . . . . . . . . . 92
3.2.10 Rack and power units configuration rules . . . . . . . . . . . . . . . . . . . . 92
3.2.11 HMC configuration rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
3.2.12 SP Cluster 1600 considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
3.2.13 Upgrade considerations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.3 IBM Configurator for e-business (e-config) . . . . . . . . . . . . . . . . . . . . . . . . 97
3.3.1 Initial order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.3.2 Performing an upgrade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
3.4 Configuration examples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.4.1 Configuration example 1: pSeries 670 (16-way 1.1 GHz) . . . . . . . . 117
3.4.2 Configuration example 2: pSeries 690 (24-way 1.3 GHz) . . . . . . . . 119
3.4.3 Model conversion from pSeries 670 to pSeries 690 . . . . . . . . . . . . 124
3.4.4 Feature conversion from POWER4 to POWER4+ . . . . . . . . . . . . . 127
iv
IBM ^ pSeries 670 and pSeries 690 System Handbook
Chapter 4. Capacity Upgrade on Demand. . . . . . . . . . . . . . . . . . . . . . . . . 133
4.1 What’s new in CUoD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.2 Description of CUoD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
4.2.1 Trial CoD function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
4.2.2 Overview of CUoD configurations . . . . . . . . . . . . . . . . . . . . . . . . . . 136
4.2.3 Supported CUoD Processor configurations . . . . . . . . . . . . . . . . . . 138
4.2.4 Supported CUoD Memory configurations . . . . . . . . . . . . . . . . . . . . 141
4.2.5 CUoD resource sequencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.2.6 Logical and physical entities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
4.2.7 CUoD license screen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
4.2.8 CUoD error messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
4.3 Activating CUoD resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.3.1 CUoD resources activation and order process . . . . . . . . . . . . . . . . 147
4.3.2 Trial CoD processor and memory . . . . . . . . . . . . . . . . . . . . . . . . . . 151
4.4 Dynamic Processor Sparing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.5 On/Off Capacity on Demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
Chapter 5. Reliability, availability, and serviceability . . . . . . . . . . . . . . . . 157
5.1 What’s new in serviceability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
5.2 RAS features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
5.3 Predictive functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.3.1 Service processor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
5.3.2 First Failure Data Capture (FFDC) . . . . . . . . . . . . . . . . . . . . . . . . . 160
5.3.3 Predictive failure analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.3.4 Component reliability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
5.3.5 Extended system testing and surveillance . . . . . . . . . . . . . . . . . . . 162
5.4 Redundancy in components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
5.4.1 Power and cooling. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
5.4.2 Memory redundancy mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . 164
5.4.3 Multiple data paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
5.5 Fault recovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
5.5.1 PCI bus parity error recovery and PCI bus deallocation . . . . . . . . . 166
5.5.2 Dynamic CPU deallocation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
5.5.3 CPU Guard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
5.5.4 Caches and memory deallocation. . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.5.5 Hot-swappable components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
5.5.6 Hot-swappable boot disks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.5.7 Hot-Plug PCI adapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
5.5.8 Light Path Diagnostics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176
5.6 Serviceability features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
5.6.1 Back up of HMC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
5.6.2 Upgrading HMC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
5.6.3 Microcode Updates function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
Contents
v
5.6.4 Inventory Scout Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
5.6.5 Service Agent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
5.6.6 Service Focal Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.6.7 Problem determination hints of Service Functions . . . . . . . . . . . . . 201
5.7 AIX RAS features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
5.7.1 Unrecoverable error analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
5.7.2 System hang detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
5.7.3 AIX disk mirroring and LVM sparing . . . . . . . . . . . . . . . . . . . . . . . . 207
5.7.4 TCP/IP RAS enhancements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208
Appendix A. Minimum and default configurations. . . . . . . . . . . . . . . . . . 209
A.1 pSeries 670 configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210
A.2 pSeries 690 configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
Appendix B. I/O loop cabling and performance . . . . . . . . . . . . . . . . . . . . 217
Abbreviations and acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
Related publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Other publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
Online resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
How to get IBM Redbooks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
vi
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figures
1-1
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
2-12
2-13
2-14
2-15
2-16
2-17
2-18
2-19
2-20
2-21
2-22
2-23
2-24
2-25
2-26
2-27
2-28
2-29
2-30
2-31
2-32
3-1
3-2
3-3
3-4
3-5
The IBM ^ pSeries 690 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
The pSeries 670 and pSeries 690 base rack with components . . . . . . . 19
CEC front view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
CEC rear view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
CEC backplane orthogonal view . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
POWER4 multichip module . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Multichip module with L2, L3, and memory . . . . . . . . . . . . . . . . . . . . . . 27
pSeries 690 1.3 GHz Turbo and 1.3 GHz HPC feature . . . . . . . . . . . . . 29
MCM, L3 cache, and memory slots relationship on backplane . . . . . . . 30
Logical relationship between MCMs, memory, and GX Slots . . . . . . . . 31
Interconnection between four MCMs . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Interconnection between processors in an MCM . . . . . . . . . . . . . . . . . . 33
MCMs and RIO ports relationship (pSeries 690) . . . . . . . . . . . . . . . . . . 41
MCM, L3 cache, and memory slots relationship on backplane . . . . . . . 43
MCMs and RIO ports relationship (pSeries 670) . . . . . . . . . . . . . . . . . . 46
Primary and secondary I/O books . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Book packaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
I/O drawer rear view. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Difference between RIO and RIO-2 connectors . . . . . . . . . . . . . . . . . . 52
Logical view of an RIO drawer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
Logical view of an RIO -G drawer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
RIO loops supported configurations. . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
RIO-2 loops supported configurations . . . . . . . . . . . . . . . . . . . . . . . . . . 59
RIO connections for three I/O drawers with IBF configuration . . . . . . . . 61
Number of usable RIO or RIO-2 ports . . . . . . . . . . . . . . . . . . . . . . . . . . 64
I/O drawer and IBF placement positions (pSeries 690) . . . . . . . . . . . . . 66
I/O drawer and IBF placement positions (pSeries 670) . . . . . . . . . . . . . 67
Physical location in frame 1 of pSeries 670 and pSeries 690 . . . . . . . . 68
Physical location in frame 2 of pSeries 690 . . . . . . . . . . . . . . . . . . . . . . 69
Media drawer power and SCSI connection . . . . . . . . . . . . . . . . . . . . . . 71
Media drawer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Power subsystem locations in BPA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Graphical user interface on the HMC. . . . . . . . . . . . . . . . . . . . . . . . . . . 76
IBM Configurator for e-business version . . . . . . . . . . . . . . . . . . . . . . . . 80
Add Initial Order icon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
The e-config main panel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Graphical representation of pSeries 690 in configuration view . . . . . . 100
pSeries 690 CEC wizard after double-clicking Rack Server 1 . . . . . . . 101
© Copyright IBM Corp. 2003. All rights reserved.
vii
3-6
3-7
3-8
3-9
3-10
3-11
3-12
3-13
3-14
3-15
3-16
3-17
3-18
3-19
3-20
3-21
3-22
3-23
3-24
3-25
3-26
3-27
3-28
3-29
3-30
4-1
4-2
4-3
4-4
4-5
4-6
4-7
4-8
4-9
4-10
4-11
5-1
5-2
5-3
5-4
5-5
5-6
5-7
viii
System tab for the pSeries 690 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Detailed diagram for the pSeries 690 CEC . . . . . . . . . . . . . . . . . . . . . 105
Detailed diagram for the media drawer . . . . . . . . . . . . . . . . . . . . . . . . 106
Selection of I/O drawer storage and RIO cabling options . . . . . . . . . . 107
Adapter tab for the I/O drawer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
Adapter tab with LAN adapter category selected. . . . . . . . . . . . . . . . . 109
Detailed diagram of the I/O drawer . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
Duplicating an existing drawer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Selecting a new I/O drawer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
HMC wizard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
HMC adapter options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
The error, warning, and information output of the configuration . . . . . 115
Right-click the error message text and it opens a menu . . . . . . . . . . . 115
Detailed error message . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Perform Final Validation icon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Add Upgrade/MES or Restore CFR icon . . . . . . . . . . . . . . . . . . . . . . . 117
Perform Final Validation icon . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
A graphical representation of the final configuration . . . . . . . . . . . . . . 123
Starting to upgrade an existing configuration. . . . . . . . . . . . . . . . . . . . 124
Start upgrading the CEC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
Changing model to pSeries 690. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Errors in the Messages tab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Changing the number of AIX licenses . . . . . . . . . . . . . . . . . . . . . . . . . 127
Duplicating I/O drawer 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Messages view during feature upgrade . . . . . . . . . . . . . . . . . . . . . . . . 130
Click to Accept on the HMC at initial boot . . . . . . . . . . . . . . . . . . . . . . 144
Click to Accept window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
CUoD menu on the HMC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Processor Capacity Settings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Save Processor Order Information . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
Send CUoD and system data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Activate CUoD Resources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
To activate the resources using Trial CoD function . . . . . . . . . . . . . . . 152
CUoD Trial CoD Processor Capacity (Next) screen . . . . . . . . . . . . . . 152
CUoD Trial CoD Processor capacity (Finish) screen . . . . . . . . . . . . . . 153
Warning message to Disable Processor Trial CoD Capacity. . . . . . . . 154
Service processor schematic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
FFDC error checkers and fault isolation registers . . . . . . . . . . . . . . . . 161
Memory error recovery mechanisms . . . . . . . . . . . . . . . . . . . . . . . . . . 164
CPU Guard activities handled by AIX . . . . . . . . . . . . . . . . . . . . . . . . . 169
CPU Guard activities handled by firmware . . . . . . . . . . . . . . . . . . . . . 170
Blind swap cassette . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Hot-Plug PCI adapters, blind swap cassette, and status LEDs . . . . . . 176
IBM ^ pSeries 670 and pSeries 690 System Handbook
5-8
5-9
5-10
5-11
5-12
5-13
5-14
5-15
5-16
5-17
5-18
5-19
5-20
5-21
5-22
5-23
5-24
5-25
5-26
5-27
5-28
5-29
5-30
5-31
5-32
B-1
B-2
B-3
B-4
B-5
B-6
B-7
B-8
Status LEDs for components in I/O drawers . . . . . . . . . . . . . . . . . . . . 177
Software maintenance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
Service Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Error reporting and consolidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
Software Maintenance: HMC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Format Media . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
Backup Critical Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Install Corrective Service . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
Software Maintenance: HMC (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
Mechanism of the Microcode Updates. . . . . . . . . . . . . . . . . . . . . . . . . 185
Microcode Updates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
Download and Apply Microcode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
Select Repository Location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
Microcode Survey Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
Microcode Installation - Device Installation . . . . . . . . . . . . . . . . . . . . . 188
Warning message . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Updating microcodes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
Microcode Updates completed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
Inventory Scout Services . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
Inventory Scout Configuration Assistant . . . . . . . . . . . . . . . . . . . . . . . 195
Service Agent on the HMC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
Service Agent on the HMC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Service Focal Point: Hardware Service Functions . . . . . . . . . . . . . . . . 199
Hardware Service Functions overview. . . . . . . . . . . . . . . . . . . . . . . . . 200
FRU LED Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200
RIO drawer, in single-loop mode on a 1.3 GHz server . . . . . . . . . . . . 218
RIO-2 drawer, in dual-loop mode on a 1.7GHz server. . . . . . . . . . . . . 219
RIO-2 drawer, in single-loop mode on a 1.7 GHz server . . . . . . . . . . . 220
Mixed RIO/RIO-2 drawer, in dual-loop mode on a 1.7 GHz server . . . 221
Migrated single-loop mode on a 1.7 GHz server . . . . . . . . . . . . . . . . . 222
Configuration for maximum bandwidth . . . . . . . . . . . . . . . . . . . . . . . . 223
Configuration for maximum number of adapters and disks . . . . . . . . . 224
Configuration after migration to RIO-2 Books without recabling. . . . . . 225
Figures
ix
x
IBM ^ pSeries 670 and pSeries 690 System Handbook
Tables
1-1
1-2
2-1
2-2
2-3
2-4
2-5
2-6
2-7
2-8
2-9
2-10
2-11
3-1
4-1
4-2
4-3
4-4
4-5
5-1
5-2
Differences between pSeries 670 and pSeries 690 . . . . . . . . . . . . . . . . 10
Comparison of AIX and Linux support for RAS features . . . . . . . . . . . . 14
pSeries 690 main subsystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Supported combinations of processors . . . . . . . . . . . . . . . . . . . . . . . . . 23
Relationship between MCMs, L3 cache, memory slots, and size . . . . . 35
Supported memory cards configurations . . . . . . . . . . . . . . . . . . . . . . . . 36
Relationship between MCMs, RIO ports, and I/O drawers . . . . . . . . . . 42
Supported memory configurations for pSeries 670 . . . . . . . . . . . . . . . . 44
Relationship between MCMs, RIO ports, and I/O drawers . . . . . . . . . . 45
Maximum configuration for each cabling pattern . . . . . . . . . . . . . . . . . . 60
Physical I/O book ports and system logical port numbers . . . . . . . . . . . 62
Available I/O ports versus installed MCM and Books . . . . . . . . . . . . . . 63
Physical location code of drawer position number . . . . . . . . . . . . . . . . . 70
Physical memory size and number of allocatable partitions . . . . . . . . . 87
CUoD processor feature codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
Supported 1.1 and 1.3 GHz CUoD processor combinations . . . . . . . . 139
Supported 1.5 and 1.7 GHz CUoD processor combinations . . . . . . . . 140
CUoD memory feature codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
CUoD error codes and messages . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
Hot-swappable FRUs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
Authentication process. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
© Copyright IBM Corp. 2003. All rights reserved.
xi
xii
IBM ^ pSeries 670 and pSeries 690 System Handbook
Notices
This information was developed for products and services offered in the U.S.A.
IBM may not offer the products, services, or features discussed in this document in other countries. Consult
your local IBM representative for information on the products and services currently available in your area.
Any reference to an IBM product, program, or service is not intended to state or imply that only that IBM
product, program, or service may be used. Any functionally equivalent product, program, or service that
does not infringe any IBM intellectual property right may be used instead. However, it is the user's
responsibility to evaluate and verify the operation of any non-IBM product, program, or service.
IBM may have patents or pending patent applications covering subject matter described in this document.
The furnishing of this document does not give you any license to these patents. You can send license
inquiries, in writing, to:
IBM Director of Licensing, IBM Corporation, North Castle Drive Armonk, NY 10504-1785 U.S.A.
The following paragraph does not apply to the United Kingdom or any other country where such
provisions are inconsistent with local law: INTERNATIONAL BUSINESS MACHINES CORPORATION
PROVIDES THIS PUBLICATION "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESS OR
IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF NON-INFRINGEMENT,
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Some states do not allow disclaimer
of express or implied warranties in certain transactions, therefore, this statement may not apply to you.
This information could include technical inaccuracies or typographical errors. Changes are periodically made
to the information herein; these changes will be incorporated in new editions of the publication. IBM may
make improvements and/or changes in the product(s) and/or the program(s) described in this publication at
any time without notice.
Any references in this information to non-IBM Web sites are provided for convenience only and do not in any
manner serve as an endorsement of those Web sites. The materials at those Web sites are not part of the
materials for this IBM product and use of those Web sites is at your own risk.
IBM may use or distribute any of the information you supply in any way it believes appropriate without
incurring any obligation to you.
Information concerning non-IBM products was obtained from the suppliers of those products, their published
announcements or other publicly available sources. IBM has not tested those products and cannot confirm
the accuracy of performance, compatibility or any other claims related to non-IBM products. Questions on
the capabilities of non-IBM products should be addressed to the suppliers of those products.
This information contains examples of data and reports used in daily business operations. To illustrate them
as completely as possible, the examples include the names of individuals, companies, brands, and products.
All of these names are fictitious and any similarity to the names and addresses used by an actual business
enterprise is entirely coincidental.
COPYRIGHT LICENSE:
This information contains sample application programs in source language, which illustrates programming
techniques on various operating platforms. You may copy, modify, and distribute these sample programs in
any form without payment to IBM, for the purposes of developing, using, marketing or distributing application
programs conforming to the application programming interface for the operating platform for which the
sample programs are written. These examples have not been thoroughly tested under all conditions. IBM,
therefore, cannot guarantee or imply reliability, serviceability, or function of these programs. You may copy,
modify, and distribute these sample programs in any form without payment to IBM for the purposes of
developing, using, marketing, or distributing application programs conforming to IBM's application
programming interfaces.
© Copyright IBM Corp. 2003. All rights reserved.
xiii
Trademarks
The following terms are trademarks of the International Business Machines Corporation in the United States,
other countries, or both:
AIX®
AIX 5L™
AS/400®
Chipkill™
ESCON®
eServer™
^™
™
IBM®
ibm.com®
IBMLink™
iSeries™
POWER2™
POWER4™
POWER4+™
POWERparallel®
PowerPC®
pSeries™
Redbooks(logo)™
Redbooks™
RS/6000®
S/370™
S/390®
SP™
zSeries®
The following terms are trademarks of other companies:
ActionMedia, LANDesk, MMX, Pentium and ProShare are trademarks of Intel Corporation in the United
States, other countries, or both.
Microsoft, Windows, Windows NT, and the Windows logo are trademarks of Microsoft Corporation in the
United States, other countries, or both.
Java and all Java-based trademarks and logos are trademarks or registered trademarks of Sun
Microsystems, Inc. in the United States, other countries, or both.
C-bus is a trademark of Corollary, Inc. in the United States, other countries, or both.
UNIX is a registered trademark of The Open Group in the United States and other countries.
SET, SET Secure Electronic Transaction, and the SET Logo are trademarks owned by SET Secure
Electronic Transaction LLC.
Other company, product, and service names may be trademarks or service marks of others.
xiv
IBM ^ pSeries 670 and pSeries 690 System Handbook
Preface
This IBM® Redbook explains the IBM ~™ pSeries™ 670 Model 671 and
the IBM ^ pSeries 690, a new level of UNIX servers providing world-class
performance, availability, and flexibility. Capable of data center, application, and
high performance computing, these servers include mainframe-inspired
self-management and security to meet your most demanding needs.
This redbook includes the following topics:
򐂰
򐂰
򐂰
򐂰
򐂰
Overview of the pSeries 670 and pSeries 690
Hardware architecture of the pSeries 670 and pSeries 690
Using the IBM Configurator for e-business
Capacity Upgrade on Demand
Reliability, availability, and serviceability
This book is an ideal desk-side reference for IBM professionals, Business
Partners, and technical specialists who support the pSeries 670 and pSeries
690, and for those who want to learn more about this radically new server in a
clear, single-source handbook. It provides the necessary information to
successfully order a pSeries 670 or pSeries 690 for a production environment
and then, upon successful installation, configure service support functions, such
as Service Focal Point and Inventory Scout.
The team that wrote this redbook
This IBM Redbook was produced by a team of specialists from around the world
working at the International Technical Support Organization, Austin Center.
Keigo Matsubara is an Advisory IT Specialist at the International Technical
Support Organization (ITSO), Austin Center. Before joining the ITSO, he worked
in the System and Web Solution Center in Japan as a Field Technical Support
Specialist (FTSS) for pSeries. He has been working for IBM for ten years.
Alvin Hua Juay Teng is a Technical Sales Specialist at IBM Singapore. He
joined IBM in 2000 and has been working with pSeries ever since. He has
experience in providing solutions for the banking industries and government
organizations. He holds a master’s degree in High Performance Computing for
Engineered Systems from Singapore-MIT Alliance (A jointed master’s degree
from Singapore University of Singapore, Singapore, Nanyang Technological
© Copyright IBM Corp. 2003. All rights reserved.
xv
University, Singapore, and Massachusetts Institute of Technology, US) and a
B.Eng in Mechanical Engineering from University of Sheffield, UK.
Bruno Blanchard is a Certified IT Specialist working for IBM France at the IGS
Pan-EMEA Infrastructure and Technology Group, in La Gaude. He holds an
Engineering degree from Ecole Centrale de Paris and a master’s degree from
Oregon State University. He has been with IBM since 1983, as a System
Engineer for VM and AIX®. He is a certified AIX and SP™ Specialist, and his
areas of expertise also include pSeries servers and clusters, as well as network
management. He is currently working as an Architect on projects deployed in
Europe for the IBM ^ cluster 1600 and pSeries servers required for the
IBM IT infrastructure.
Thanks to the following people for their contributions to this project:
International Technical Support Organization, Austin Center
Scott Veter and Wade Wallace
International Technical Support Organization, Rochester Center
Gabrielle Velez
IBM Austin
Alan Standridge, Andy McLaughlin, Bill Casey, Bob Foster, Bruce Wood, Dave
Willoughby, Doug Bossen, Eric Marshall, Iggy Haider, James Donnelly, Joel M.
Tendler, John Bissell, John Purcell, Kaena Freitas, Mark Dewalt, Mike Stys, Minh
Nguyen, Richard Bolton, Truc Nguyen, Walter Lipp
IBM Endicott
Scott Nettleship and Lenny Nichols
IBM Poughkeepsie
James T Mitchell
IBM Rochester
David L Shaw, Lawny Miller, Matt Spinler, Sandy Shirk-Heath
Become a published author
Join us for a two- to six-week residency program! Help write an IBM Redbook
dealing with specific products or solutions, while getting hands-on experience
with leading-edge technologies. You will team with IBM technical professionals,
Business Partners and/or customers.
xvi
IBM ^ pSeries 670 and pSeries 690 System Handbook
Your efforts will help increase product acceptance and customer satisfaction. As
a bonus, you'll develop a network of contacts in IBM development labs, and
increase your productivity and marketability.
Find out more about the residency program, browse the residency index, and
apply online at:
ibm.com/redbooks/residencies.html
Comments welcome
Your comments are important to us!
We want our Redbooks™ to be as helpful as possible. Send us your comments
about this or other Redbooks in one of the following ways:
򐂰 Use the online Contact us review redbook form found at:
ibm.com/redbooks
򐂰 Send your comments in an Internet note to:
[email protected]
򐂰 Mail your comments to:
IBM Corporation, International Technical Support Organization
Dept. JN9B Building 003 Internal Zip 2834
11400 Burnet Road
Austin, Texas 78758-3493
Preface
xvii
xviii
IBM ^ pSeries 670 and pSeries 690 System Handbook
Summary of changes
This section describes the technical changes made in this edition of the book and
in previous editions. This edition may also include minor corrections and editorial
changes that are not identified.
Summary of Changes
for SG24-7040-02
for IBM ^ pSeries 670 and pSeries 690 System Handbook
as created or updated on March 29, 2005.
April 2003, Third Edition
This revision reflects the addition, deletion, or modification of new and changed
information described below.
New information
򐂰
򐂰
򐂰
򐂰
򐂰
򐂰
򐂰
AIX 5L Version 5.2
Linux information on the pSeries 670 and pSeries 690
POWER4™+ processors
New I/O subsystem
Capacity Upgrade on Demand for memory
How to update HMC
Microcode update service
Changed information
򐂰 Using the IBM Configurator for e-business
򐂰 Capacity Upgrade on Demand for processors
򐂰 Description of On/Off Capacity on Demand
October 2002, Second Edition
The second version of this book, SG24-7040-01 IBM ^ pSeries 670 and
pSeries 690System Handbook, was written by the following authors:
Keigo Matsubara, Bas Hazelzet, Marc-Eric Kahle
This revision reflects the addition, deletion, or modification of new and changed
information described below.
© Copyright IBM Corp. 2003. All rights reserved.
xix
New information
򐂰 The IBM ^ pSeries 670 Model 671
򐂰 Capacity Upgrade on Demand
򐂰 Overview of the feature codes
Changed information
򐂰 Detailed logical partitioning explanation was moved to The Complete
Partitioning Guide for IBM ^ pSeries Servers, SG24-7039.
򐂰 Usage examples about how to use IBM Configurator for e-business were
updated to include the IBM ^ pSeries 670 Model 671.
򐂰 A new explanation was added about service functions, such as Service Focal
Point and Inventory Scout.
򐂰 The CUoD support for AIX 5L Version 5.1 was added.
The following list shows contributors for the first version of this book,
SG24-7040-00 IBM ^ pSeries 690 System Handbook:
IBM Austin
Alan Standridge, Andy McLaughlin, Bob Foster, Bruce Wood, Dave Lewis, Dave
Willoughby, Doug Bossen, George W Clark, James Donnelly, Jane Chilton, John
O'Quin, John Purcell, Larry Amy, Luke Browning, Mark Dewalt, Mike Stys, Minh
Nguyen, Sajan Lukose, Siraj Ismail, Susan Caunt, Truc Nguyen, Walter Lipp
IBM Endicott
Scott Nettleship and Lenny Nichols
IBM Germany
Knut Muennich
IBM Hamden
Jack Dobkowski
IBM Poughkeepsie
Anthony Pioli, Juan Parrilla II, Michael Schmidt, Rob Overton
IBM Rochester
Matt Spinler
IBM UK
Dave Williams and Clive Benjamin
xx
IBM ^ pSeries 670 and pSeries 690 System Handbook
March 2002, First Edition
The first version of this book, SG24-7040-00 IBM ^ pSeries 690 System
Handbook, was written by the following authors:
Keigo Matsubara, Cesar Diniz Maciel, KangJin Lee, Martin Springer
The following list shows contributors for the first version of this book,
SG24-7040-00 IBM ^ pSeries 690 System Handbook:
IBM Austin
Alan Standridge, Andy McLaughlin, Arthur Ban, Balaram Sinharoy, Bob Foster,
Bradley McCredie, Dave Willoughby, David Ruth, Doug Bossen, Duke Paulsen,
Eric Marshall, Gerald McBrearty, Hung Le, Hye-Young McCreary, Jan Klockow,
Joel M. Tendler, Julie Craft, Kurt Szabo, Luc Smolders, Mike Stys, Pat Buckland,
Randy Itskin, Richard Cutler, Robert W West, Steve Dodson, Steve Fields,
Steven Hartman, Susan Caunt, Walter Lipp, William Hodges
IBM Japan
Shigeru Kishikawa and Shingo Matsuda
IBM Poughkeepsie
Ron Goering
Summary of changes
xxi
xxii
IBM ^ pSeries 670 and pSeries 690 System Handbook
1
Chapter 1.
Overview of the pSeries 670
and pSeries 690
The IBM ^ pSeries 670 Model 671 (hereafter referred to as pSeries 670)
and the IBM ^ pSeries 690 (hereafter referred to as pSeries 690) are
integral parts of the IBM ^ product line, featuring servers that can help
lower total cost of ownership, improve efficiency, and speed up e-business
transformation.
The pSeries 670 and pSeries 690 are IBM’s flagship products in the pSeries line
of UNIX servers. The pSeries 670, a 4-way to 16-way server, and the pSeries
690, an 8-way to 32-way server, share the same basic technology, and represent
the latest generation of performance leadership and reliability from IBM. They
incorporate the latest advances in chip technology from IBM, as well as many
leading self-managing system capabilities to enable mission-critical operation.
In addition to unparalleled speed, the pSeries 670 and pSeries 690 systems
have the ability to consolidate critical applications on a single, data center class
server. As a result, there are fewer servers to manage and maintain. Available
capacity can be used more effectively and with greater flexibility by the dynamic
logical partitioning function to meet changing business demands.
In this chapter we introduce the pSeries 670 and pSeries 690 servers,
highlighting their advantages, main characteristics and components, as well as
how they fit into the data center.
© Copyright IBM Corp. 2003. All rights reserved.
1
1.1 What’s new in the pSeries 670 and pSeries 690
This third edition of the IBM ^ pSeries 670 and pSeries 690 System
Handbook, SG24-7040 describes these two models of servers, as available after
the products’ announcements in May 2003, that brings the following new
features1:
򐂰 Processors using the POWER4+ chip, at 1.5 and 1.7 GHz
򐂰 L3 Cache and memory modules running at 500 and 575 MHz
򐂰 64 GBs memory modules
򐂰 A faster I/O subsystem, using new I/O books, new RIO cables, new I/O drawer
planar
򐂰 Support for 133 MHz PCI-X
򐂰 Improved partitioning facility, with support up to 32 partitions
򐂰 Improved Capacity Upgrade on Demand addressing both the processor and
the memory resources
򐂰 new upgrades possibilities, including model conversion from pSeries 670 to
pSeries 690
򐂰 Firmware updates from the HMC
The support of these new features leaves the general architecture of the pSeries
670 and pSeries 690 servers unchanged. However, it results into many important
implementation changes, such as the RIO drawer cabling, the PCI adapters
supported combinations, or the manner to use the IBM Configurator for
e-business to order a new server or an MES (Miscellaneous Equipment
Specification).
If you are already familiar with the pSeries 670 and pSeries 690 servers, you can
avoid reading their general presentation in the following sections, and directly
skip to the following chapters.
If however you are new to the pSeries 670 and pSeries 690 servers, we strongly
encourage you to read this introductory chapter in its entirety.
Note: Throughout this book, we use the word “POWER4” to refer to the 1.1,
1.3, 1.5 or 1.7 GHz processors when the technology or speed difference is not
of importance, and we specify “POWER4+” to refer exclusively to the 1.5 and
1.7 GHz processors.
1
2
Please check the announcement letter for the exact availability date of each new feature.
IBM ^ pSeries 670 and pSeries 690 System Handbook
1.2 pSeries 670 and pSeries 690 characteristics
The pSeries 670 and pSeries 690 family of servers incorporate advanced
technologies available from across the IBM ^ family, as well as
technology enhancements from IBM research divisions. The result are
high-performance, high-availability servers that offer enhanced features and
benefits.
The pSeries 670 and pSeries 690 servers are based on a modular design. They
feature a Central Electronics Complex (CEC) (where memory and processors
are installed), power subsystem, and I/O drawers, also known as Remote I/O
(RIO) drawers. Optional battery backups can provide energy for an emergency
shutdown in case of a power failure.
The CEC provides room for up to 16 POWER4 chips, each one housing two
processors and a shared level 2 (L2) cache. The POWER4 chip offers an
advanced microprocessor design, with an SMP design in a single silicon
substrate.
Advanced multichip module (MCM) packaging places four POWER4 chips with
four or eight processors, mounted on a ceramic package. This provides
enhanced processor reliability, and faster interconnections between processors
and L2 caches. Each MCM connects to two memory card slots through the level
3 (L3) cache modules.
Redundant power supplies and optional battery backups assure that a single
power failure does not interrupt system operation. In case of total power failure,
an emergency shutdown is initiated whenever the battery is present. The
redundant power supplies are connected through separate power cords, in order
to prevent a single failure on the external electrical circuit.
Building on IBM ^ zSeries® heritage, the pSeries 670 and pSeries 690
deliver true logical partitioning and can be divided respectively into up to 16 or 32
partitions2, each with its own set of system resources, such as processors,
memory, and I/O. Unlike partitioning techniques available on other UNIX servers,
logical partitioning provides greater flexibility and finer granularity. Resources can
be assigned in any amount or combination for business-critical applications.
Each logical partition has its own version of an operating system, being either
AIX 5L Version 5.1, AIX 5L Version 5.2, or Linux.
The following sections detail some of the technologies behind the pSeries 670
and pSeries 690.
2
The number of LPAR that can be instantiated on a system depends on its hardware configuration.
Please refer to the The Complete Partitioning Guide for IBM ^ pSeries Servers, SG24-7039
for explanation of the LPAR configuration rules, and to 2.3.6, “I/O books” on page 47 in this redbook.
Chapter 1. Overview of the pSeries 670 and pSeries 690
3
1.2.1 Microprocessor technology
The POWER4 chip is a result of advanced research technologies developed by
IBM. Numerous technologies are incorporated into the POWER4 to create a
high-performance, high-scalability chip design to power future IBM ^
pSeries systems. Some of the advanced techniques used in the design and
manufacturing processes of the POWER4 include copper interconnects and
Silicon-on-Insulator.
Copper interconnects
As chips become smaller and faster, aluminum interconnects, which have been
used in chip manufacturing for over 30 years, present increasing difficulties. In
1997, after nearly 15 years of research, IBM scientists announced a new
advance in the semiconductor process that involves replacing aluminum with
copper. Copper has less resistance than aluminum, which permits the use of
smaller circuits with reduced latency that allows for faster propagation of
electrical signals. The reduced resistance and heat output make it possible to
shrink the electronic devices even further while increasing clock speed and
performance without resorting to exotic chip cooling methods.
Silicon-on-Insulator
Silicon-on-Insulator (SOI) refers to the process of implanting oxygen into a silicon
wafer to create an insulating layer and using an annealing process until a thin
layer of SOI film is formed. The transistors are then built on top of this thin layer of
SOI. The SOI layer reduces the capacitance effects that consume energy,
generate heat, and hinder performance. SOI improves chip performance by 25 to
35 percent.
Detailed information on the POWER4 chip can be found in 2.3.1, “POWER4
processor and MCM packaging” on page 22.
1.2.2 Memory subsystem
The memory hierarchy in pSeries 670 and pSeries 690 systems is represented
as follows:
4
Internal level 1 (L1) caches
Each microprocessor core has its own L1
instruction and L1 data caches.
Shared L2 cache
Inside a POWER4 chip, both processor cores
share an L2 cache.
IBM ^ pSeries 670 and pSeries 690 System Handbook
L3 cache
It operates at a 3:1 ratio with the clock speed,
and connects to the MCMs through a dedicated
port. Each L3 cache module has a direct
connection with one POWER4 chip and one
memory slot.
The memory used in the pSeries 670 and pSeries 690 systems is Double Data
Rate (DDR), which provides superior bandwidth. The minimum memory in the
pSeries 670 system is 4 GB, in the pSeries 690 it is 8 GB, and it can be
expanded to up to 256 GB and 512 GB, respectively. The pSeries 670 and
pSeries 690 are the first pSeries servers to incorporate level 3 (L3) cache. This
contributes to faster data access and faster processing capabilities.
Additional information about memory can be found in 2.3.2, “Memory subsystem
for pSeries 690” on page 29, and in 5.4.2, “Memory redundancy mechanisms” on
page 164.
1.2.3 I/O drawer subsystem
The standard system comes with one I/O drawer that features 20 Hot-Plug PCI
slots, and supports more than 2.3 TB of storage located in 16 drive bays
connected to integrated Ultra3 SCSI controllers. All power, thermal control, and
communication systems have redundancy to eliminate outages caused by single
component failures. Each I/O drawer connects to the system through two or four
RIO ports, with up to 8 GB/s of total burst bandwidth per drawer3.
Up to three drawers can be added to the pSeries 670 to obtain a total of 60 PCI
slots and more than 7 TB of internal disk storage. The pSeries 690 can
accommodate up to eight drawers, supporting a total of 160 PCI slots and more
than 18 TB of internal disk storage. For more information on the I/O subsystem,
please refer to 2.4, “I/O subsystem” on page 51.
3
In the case of an RIO drawer connected to the CEC through two RIO loops.
Chapter 1. Overview of the pSeries 670 and pSeries 690
5
A picture of the IBM ^ pSeries 690 is shown in Figure 1-1.
Figure 1-1 The IBM ^ pSeries 690
1.2.4 Hardware technology
The pSeries 690 introduces several advanced technologies into its design, and
implements other available technologies with enhancements, truly delivering
mainframe-class reliability and an outstanding performance. Some key
technologies used are referenced here, as follows:
򐂰 Many self-healing, self-optimizing, and fault-avoidance mechanisms.
򐂰 zSeries packaging and power distribution, recognized as highly reliable.
򐂰 Advanced microprocessor design, with POWER4 dual processors, and
multichip module (MCM) packaging.
򐂰 Logical partitioning (LPAR) offers true flexibility when selecting resources for
partitions, and dynamic logical partitioning (DLPAR) allows for moving
resources non-disruptively between logical partitions.
6
IBM ^ pSeries 670 and pSeries 690 System Handbook
The pSeries 670 and pSeries 690 systems are the most reliable UNIX servers
ever built, with the pSeries 690 the most powerful UNIX machine on the market.
To achieve the outstanding capabilities that these servers offer, IBM has
combined two methods to select the technology: Developing new technology, and
enhancements to existing technology.
Developing new technologies is one of the most challenging tasks, especially in
the high-end UNIX market. It means increased risks, because of the cost of
development and the possibility of missing the product release schedule. As a
huge technological step in the high-end UNIX market, IBM developed several
new technologies for the pSeries 690, such as the POWER4 processor and
distributed switch technology. These technologies are now used across the
whole range of the pSeries Family, from the entry model pSeries 610 to the
high-end pSeries 690 and the clustered model pSeries 655. Hardware
technology is shared with the IBM ^ iSeries™.
Enhancing existing technologies is an excellent way to exploit solutions already
developed, and add new functions and features. It enables technology reuse,
lowering costs and enabling a new product to use the best technology available
for each specific requirement. The zSeries power subsystem and packaging and
the MCM design are among those technologies that have been enhanced on the
pSeries 670 and pSeries 690.
1.3 Logical partitioning
There is a strong demand for high-end systems to provide greater flexibility, in
particular the ability to subdivide them into smaller partitions that are capable of
running a version of an operating system or a specific set of application
workloads.
IBM initially started work on partitioning S/370™ mainframe systems in the
1970s. Since then, logical partitioning on IBM mainframes (now called IBM
^ zSeries) has evolved from a predominantly physical partitioning
scheme based on hardware boundaries, to one that allows for virtual and shared
resources with dynamic load balancing. In 1999 IBM implemented LPAR support
on the AS/400® (now called IBM ^ iSeries) platform. In 2000 IBM
announced the ability to run the Linux operating system in an LPAR or on top of
VM on a zSeries server, to create thousands of Linux instances on a single box.
Chapter 1. Overview of the pSeries 670 and pSeries 690
7
Throughout this publication we refer to the different partitioning mechanisms
available on the market. Therefore, it is appropriate to clarify the terms and
definitions by which we classify these mechanisms.
򐂰 A building block is a collection of system resources, such as CPUs, memory,
and I/O connections. These may be physically packaged as a self-contained
SMP system (rack-mounted or stand-alone) or as boards within a larger
multiprocessor system. There is no requirement for the CPUs, memory, and
I/O slots to occupy the same physical board within the system, although they
often do. Other vendors use the terms system board, cell, and Quad Building
Block (QBB) to refer to their building blocks.
򐂰 A physical partition consists of one or more building blocks linked together by
a high-speed interconnect. Generally, the interconnect is used to form a
single coherent memory address space. In a system that is only capable of
physical partitioning, a partition is a group of one or more building blocks
configured to support an operating system image. Other vendors may refer to
physical partitions as domains or nPartitions.
򐂰 A logical partition is a subset of logical resources that are capable of
supporting an operating system. A logical partition consists of CPUs,
memory, and I/O slots that are a subset of the pool of available resources
within a system.
Note: The major difference between logical partitioning and physical
partitioning is the granularity and flexibility available for allocating resources to
an operating system image. Logical partitions have finer granularities than
physical partitions.
It should be noted that the zSeries LPAR implementation is unique in comparison
to the other partitioning implementations available from IBM and other hardware
vendors. It is the most mature and dynamic partitioning technology in the
industry. IBM experience with physical and logical partitioning over the last 25
years has greatly influenced the design and implementation of logical partitioning
on pSeries.
The pSeries 690 were the first pSeries servers to incorporate the ability to be
partitioned (This is now also available on pSeries 630, 650, 655, and 670). Their
architectural design brings true logical partitioning to the UNIX world, being
capable of up to 32 partitions inside a single server, with great flexibility in
resource selection. The partitioning implementation on the pSeries 670 and
pSeries 690 differs from those of other UNIX system vendors in that the physical
resources that can be assigned to a partition are not limited by internal physical
system board boundaries.
8
IBM ^ pSeries 670 and pSeries 690 System Handbook
Processors, memory, and I/O slots can be allocated to any partition, regardless
of their locality. For example, two processors on the same POWER4 silicon chip
can be in different partitions. PCI slots are assigned individually to partitions, and
memory can be allocated in fixed-size increments. The granularity of the
resources that can be assigned to partitions is very fine, providing flexibility to
create systems with just the desired amount of resources.
The pSeries 670 and pSeries 690 systems can also be shipped with additional
capacity, which may be purchased and activated at a certain point in time without
affecting normal machine operation. This feature is referred to as Capacity
Upgrade on Demand (CUoD), which provides flexibility and fine granularity in
processor and memory upgrades. A feature that comes with CUoD is Dynamic
Processor Sparing. This is the capability of the system to configure out a failing
processor and configure in a non-activated CUoD processor. In this way, the
number of activated processors is guaranteed by IBM whenever a processor
failure would occur. Details on this new offering can be found in Chapter 4,
“Capacity Upgrade on Demand” on page 133.
The pSeries 670 and pSeries 690 are also capable of running Linux inside a
partition, and having both AIX 5L Version 5.1, AIX 5L Version 5.2, and Linux
running on the system simultaneously. The pSeries 690 is the first high-end
UNIX server capable of running Linux, and the only one supporting Linux
concurrently with other operating systems.
For more information on logical partitioning technology on the pSeries 690, refer
to the Partitioning for the IBM ^ pSeries 690 System white paper,
available at:
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/lpar.html
Also refer to The Complete Partitioning Guide for IBM ^ pSeries Servers,
SG24-7039.
1.4 Dynamic logical partitioning
Starting from AIX 5L™ Version 5.2, IBM supports dynamic logical partitioning
(also known as DLPAR) in partitions on several logical-partitioning capable IBM
^ pSeries server models, including the pSeries 670 and pSeries 690.
The dynamic logical partitioning function allows you to add and remove
resources, such as CPUs, memory, and I/O slots, to and from a partition as well
as to move resources between two partitions, without operating system reboot
(on the fly). However, to use the function, the following conditions must be met,
because it is achieved by the combination of these three components:
Chapter 1. Overview of the pSeries 670 and pSeries 690
9
򐂰 The pSeries 670 and pSeries 690 systems must be updated with the 10/2002
system microcode update or later.
򐂰 The IBM Hardware Management Console for pSeries (HMC) must be
updated with Release 3 or higher, and must be connected through Ethernet to
the each partition.
򐂰 A partition that requires the dynamic logical partitioning function must be
installed with AIX 5L Version 5.2, or must be migrated to AIX 5L Version 5.2
from AIX 5L Version 5.1.
1.5 General overview of pSeries 670 and pSeries 690
The pSeries 670 has some differences from the pSeries 690. Basically it has the
same technical advantages and Reliability, Availability, and Serviceability (RAS)
features, but it is a single rack server and has several configuration limitations
compared to the pSeries 690. Table 1-1 provides a quick overview of the
differences between these two models.
Table 1-1 Differences between pSeries 670 and pSeries 690
10
Component
pSeries 690
pSeries 670
Number of CPUs / MCMs
8-32/1-4
4-16/1-2
4 way 1.1GHz POWER4
No
Yes
8 way 1.1GHz POWER4
Yes
Yes
8 way 1.3GHz POWER4
Turbo
Yes
No
4 way 1.3GHz POWER4
HPC
Yes
No
4 way 1.5GHz POWER4+
No
Yes
8 way 1.5GHz POWER4+
Yes
Yes
8 way 1.7GHz POWER4+
Yes
No
Amount of memory (GB)
8-512
4-256
Number of memory cards
1-8
1-4
Number of I/O drawers
1-8
1/2a - 3
Number of PCI slots
20-160
10-60
IBF feature
Yes
Yes
IBM ^ pSeries 670 and pSeries 690 System Handbook
Component
pSeries 690
pSeries 670
Number of racks
1-2
1
a. As an initial order, just one half I/O drawer can be used with 10 PCI slots.
The MCMs used in the pSeries 670 are either 4- or 8-way modules. Since it is
only possible to install two MCMs in the pSeries 670, only the inward-facing
memory card slots can be used.
Note: It is possible to upgrade a pSeries 670 into a pSeries 690 by model
conversion.
1.6 Market positioning
The pSeries 690 is one of the most powerful UNIX servers ever built, with the
best performance per processor in many application areas. The same holds for
the pSeries 670 in the midrange. From copper-based, Silicon-on-Insulator (SOI)
POWER4 microprocessor technology and multichip packaging expertise to the
industry’s most advanced self-management capabilities, the pSeries 670 and
pSeries 690 are designed to provide the ultimate in performance, configuration
flexibility, and availability.
The pSeries 670 and pSeries 690 are positioned to fulfill the applications’ need
for performance and scalability, and the reliability required in mission-critical
environments. It is an ideal server on which to run corporate applications such as
Online Transaction Processing (OLTP), Enterprise Resource Planning (ERP),
Business Intelligence (BI), and high-performance e-business infrastructures. It
also provides attractive facilities for server consolidation with its flexible
partitioning mechanism, as well as the dynamic logical partitioning function,
possibilities, and advanced management functions.
Performance wise, the pSeries 670 and pSeries 690 excel in commercial
processing, and the pSeries 690 set many new records in High Performance
Computing (HPC) applications. The balanced architecture, together with high
performance processors, offer unparalleled price/performance, helping to reduce
costs not only on the server itself, but also on power, cooling, and software
licenses.
In the case of very resource-consuming applications, or for consolidation of a
very high number of servers, it is possible to group several pSeries 670 and
pSeries 690 into clusters, (optionally with other pSeries, RS/6000® and SP
servers) known as IBM ^Cluster 1600.
Chapter 1. Overview of the pSeries 670 and pSeries 690
11
Additional clustering information can be found at:
http://www.ibm.com/servers/eservers/clusters/hardware/1600.html
For more information on the AIX 5L operating system and related IBM products,
the following link may be of interest:
http://www.ibm.com/servers/aix/
For application availability on the AIX 5L operating system, the following link can
be used for alphabetical listing and advanced search options, both for IBM
software products and third-party software products.
http://www.ibm.com/servers/aix/products/
1.7 Supported operating systems
The goal of this redbook is to present the hardware features of the pSeries 670
and pSeries 690 servers. It does not have a designated chapter providing
information about the operating systems that are supported on this hardware.
However, throughout the book, when a feature requires a specific level of
software, we will mention the prerequisite.
The following sections provide general information and pointers to the relevant
documentation about operating systems you can run on these servers:
򐂰 “AIX 5L Version 5.1” on page 12
򐂰 “AIX 5L Version 5.2” on page 13
򐂰 “Linux - SuSE” on page 13
1.7.1 AIX 5L Version 5.1
When using AIX on pSeries 670 or pSeries 690, AIX 5L Version 5.1 is the
minimum level that must be installed. It is the first version of AIX supporting the
POWER 4 processor technology. Any prior version of AIX, including AIX 4.3.3, is
not supported on pSeries 670 or pSeries 690.
AIX 5.1 provides support for the features that were announced on the first
generation of pSeries 690. It therefore does not support some of the features that
were announced later. In particular, AIX 5.1 does not support:
򐂰
򐂰
򐂰
򐂰
12
Dynamic LPAR
Memory Capacity Upgrade on Demand
Dynamic Processor Sparing
Dynamic CPU Guard
IBM ^ pSeries 670 and pSeries 690 System Handbook
While models that support greater than 256 GB memory sizes have the potential
to define greater than 256 GB logical partition sizes, AIX 5.1 logical partitions
have a maximum logical partition memory size of 256 GB. AIX 5.1 partitions
defined to have greater than 256 GB would fail to activate with an insufficient real
mode memory failure. AIX 5.2 and Linux partitions can be greater than 256 GB in
size.
1.7.2 AIX 5L Version 5.2
AIX 5L Version 5.2 is the latest available version of AIX. As a general rule, all
features available with the pSeries 670 or pSeries 690 are supported by this level
of AIX. Some features may require a specific maintenance level.
1.7.3 Linux - SuSE
The Linux distribution which is supported on the pSeries 670 or pSeries 690 is
the SuSE Linux Enterprise Server 8. Other distributions of Linux are not officially
supported on pSeries 670 or pSeries 690, although some distributors have
announced their intention to port their products to this hardware platform.
To find the latest information about the Linux products supported on pSeries,
refer to these Web sites:
http://www.ibm.com/servers/eserver/pseries/linux/
http://www.ibm.com/servers/eserver/pseries/linux/whitepapers/linux_pseries.pdf
Information about SuSE Linux Enterprise Server 8 can be found at:
http://www.suse.com/us/business/products/server/sles/i_pseries.html
Release notes for the SuSE Linux Enterprise Server 8 are available at:
http://www.ibm.com/servers/eserver/pseries/linux/sles8_release_notes.pdf
There are some restrictions that you must be aware of regarding Linux on
pSeries 670 or pSeries 690:
򐂰 Linux-SuSE can only be installed on partitions and not supported on the Full
System Partition.
򐂰 Only a subset of the hardware options (disks, adapters, ...) are supported. For
the latest list of supported features, refer to:
http://www.ibm.com/servers/eserver/pseries/hardware/linux_facts.pdf
򐂰 We do not recommend using Linux in partition with a large number of
processors. Usually, Linux applications perform at their best with less than 4
to 8 processors in the partition.
򐂰 CuOD is not available for Linux systems.
Chapter 1. Overview of the pSeries 670 and pSeries 690
13
Redbooks are available that specifically address the topic of Linux on pSeries. In
particular, we recommend that you read Linux Applications on pSeries,
SA24-6033, for information about installing Linux on pSeries 670 or pSeries 690,
or porting Linux applications onto AIX using the AIX toolbox for Linux
applications.
1.7.4 Comparison of RAS supported features
The hardware and firmware of pSeries 670 or pSeries 690 provides for many
RAS features, which are described in Chapter 5, “Reliability, availability, and
serviceability” on page 157.
Table 1-2 Comparison of AIX and Linux support for RAS features
14
RAS functions
AIX
Linux
Service Focal Point
Yes
No
I/O Errors Reporting
Yes
Yes
EEH Detection
Yes
Yes
EEH Recovery
Yes
No
Concurrent Diagnostics
Yes
No
Hot Plug (I/O, SCSI)
Yes
No
Inventory Scout
Yes
No
Update System Firmware
Yes
Yes
Update IOA Firmware
Yes
No
Platform errors to syslog
Yes
Yes
EPOW Warnings
Yes
No
lscfg
Yes
Partial
Scan Dump
Yes
Yes
DASD Mirroring
Yes
No
Service Agent
Yes
No
Error Log Analysis
Yes
No
Predictive callout
Yes
No
Dynamic CPU Deallocation
Yes
No
snap
Yes
Yes
IBM ^ pSeries 670 and pSeries 690 System Handbook
RAS functions
AIX
Linux
lsvpd
Yes
Yes
lsmcode
Yes
Yes
SRC codes(IPL prog, Machine check, dump state
Yes
No
HACMP Options
Yes
No
The RAS features also need some support by the operating system, so that the
system administrator can use them. Table 1-2 on page 14 lists which features are
available in AIX or Linux.
Some features may require a specific maintenance level of code to be supported.
For many features that are not supported by Linux, there are statements of the
direction for support in future versions of the operating system.
1.7.5 Installation and backup of the operating systems
For a pSeries 670 or pSeries 690, all internal media bays on the front are
connected to one I/O slot, while all internal media bays in the back are connected
to another I/O slot. Therefore, a maximum of two partitions can be directly
connected to a set of internal media bays at any one time. In many cases,
because of the performance and flexibility of the systems, more than two
partitions are configured on a pSeries 670 or pSeries 690. In this section, we
briefly describe the three following options to install and back up multiple
partitions on a pSeries 670 or pSeries 690.
򐂰 Share over time a set of internal media (CD-ROM and tape drive) between all
the partitions. This option is likely to have the lowest cost and works with both
AIX and Linux. But note that unless you are running AIX 5L Version 5.2, you
would need to shutdown the partitions that are affected when transferring the
media bays from one partition to another. Another point to take into account is
that at any one time, a maximum of two partitions can be installed or backed
up if you have the maximum number of internal media bays.
򐂰 Provide a set of media (CD-ROM and tape drive) for each partition. If there is
more than two partitions, you need to configure additional external CD-ROM,
tape drives and the required adapters on the I/O drawers. This option can be
used for both AIX and Linux. The advantage if this option is that all the
partitions can be scheduled to be backed up at the same time or different
times, but there is an additional cost for the external devices.
򐂰 Provide a set of internal media to a partition and use the AIX Network
Installation Manager (NIM) to install and manage the partitions. This option is
only supported with AIX. NIM provides capabilities, such as installation, fixes
Chapter 1. Overview of the pSeries 670 and pSeries 690
15
through the network, and also creates system backups using the mksysb
command. For more details about NIM, refer to The Complete Partitioning
Guide for IBM ^ pSeries Servers, SG24-7039.
16
IBM ^ pSeries 670 and pSeries 690 System Handbook
2
Chapter 2.
Hardware architecture of the
pSeries 670 and pSeries 690
Both the pSeries 670 and pSeries 690 are based on a modular design, where all
components are mounted in one or two 24 inch racks. There are three major
subsystems: the Central Electronics Complex (CEC), the power subsystem, and
the I/O subsystem. In addition, all these components are managed by an
external management workstation, called the Hardware Management Console
(HMC). Each of these subsystems and the HMC are explained in detail in this
chapter:
򐂰
򐂰
򐂰
򐂰
2.3, “Central Electronics Complex” on page 19
2.4, “I/O subsystem” on page 51
2.5, “Power subsystem” on page 72
2.6, “IBM Hardware Management Console for pSeries” on page 75
The pertinent information about the differences between the pSeries 690 and the
pSeries 670 is also provided in this chapter.
The support of the Capacity Upgrade on Demand feature on the pSeries 670 and
pSeries 690 servers requires some specific hardware features and introduces
additional configuration rules. All CUoD specific aspects are described in
Chapter 4, “Capacity Upgrade on Demand” on page 133, and are not addressed
in this chapter.
© Copyright IBM Corp. 2003. All rights reserved.
17
2.1 What’s new in the pSeries 670 and pSeries 690
In May 2003, IBM has announced new hardware features on the pSeries 690 and
pSeries 670 servers:
򐂰 In addition to the 1.1GHz and 1.3 GHz chip using POWER4 technology, the
customer can now also choose processors using the POWER4+ technology,
with a clock speed of 1.5 or 1.7 GHz. These new processors are described in
2.3.1, “POWER4 processor and MCM packaging” on page 22, as well as their
new associated L3 cache modules.
򐂰 As for the L3 cache, new memory modules are announced, that support a
faster clock speed to be used with the new POWER4+ processors. In
addition, a larger 64 GB memory board is now supported. These memory
modules and the new configuration possibility they offer are described in
2.3.2, “Memory subsystem for pSeries 690” on page 29 and 2.3.4, “Memory
subsystem for pSeries 670” on page 42.
򐂰 A completely new I/O subsystem, using a faster technology (called RIO-2), is
mandatory to support the new POWER4+ processors, and can also be used
in conjunction with the POWER4 processors. It consists of new I/O books,
described in 2.3.6, “I/O books” on page 47, new I/O planers for the RIO
drawers (see 2.4.1, “I/O drawer” on page 51), and a completely new set of
cabling options between the I/O books and the RIO drawers, explained in
2.4.2, “I/O subsystem communication and monitoring” on page 56.
2.2 Modular design of the pSeries 670 and pSeries 690
Both the pSeries 670 and pSeries 690 are rack-based servers, housed on the
same 24-inch wide, 42 EIA height rack used by the IBM ^ zSeries. Inside
this rack all the server components are placed in specific positions. This design
and mechanical organization offers advantages in optimization of floor space
usage.
Table 2-1 shows the main subsystems of the pSeries 690 and pSeries 670, and
the sections where they are described.
Table 2-1 pSeries 690 main subsystems
18
Subsystem name
Section number
Central Electronics Complex (CEC)
2.3, “Central Electronics Complex” on page 19
I/O subsystem
2.4, “I/O subsystem” on page 51
2.4.4, “Media drawer” on page 70
Power subsystem
2.5, “Power subsystem” on page 72
IBM ^ pSeries 670 and pSeries 690 System Handbook
A representation of these components inside the base rack is shown in
Figure 2-1.
24-Inch System Rack, 42 EIA
Bulk Power Subsystem
(integrated in rack)
8 EIA
17 EIA
Central Electronics Complex (CEC)
IBM Hardware
Management
Console for 1 EIA
pSeries
2 EIA
(HMC)
Media Drawer (required)
Up to four optional Internal Batteries Features
(This space can be used for a fourth I/O drawer
for pSeries 690 only)
2 EIA
First I/O Drawer (required)
4 EIA
pSeries 670 supports 1/2 I/O drawer (10 PCI slots)
4 EIA
Optional I/O Drawer
4 EIA
Optional I/O Drawer
Figure 2-1 The pSeries 670 and pSeries 690 base rack with components
2.3 Central Electronics Complex
The Central Electronics Complex is a 17 EIA height drawer housing the
processors and memory of the pSeries 670 and pSeries 690. The CEC contains
the following components:
򐂰 The CEC backplane, where the components are mounted
򐂰 The multichip modules (MCMs), which contain the POWER4 processors
򐂰 Memory cards
򐂰 L3 cache modules
򐂰 I/O books, which provide the Remote I/O (RIO) ports for the I/O drawers, and
the service processor function
򐂰 Fans and blowers for CEC cooling
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
19
Major design efforts have contributed to the development of the pSeries 670 and
pSeries 690 to analyze single points of failure within the CEC to either eliminate
them or to provide hardening capabilities to significantly reduce their probability
of failure.
The front view of CEC is shown in Figure 2-2 on page 20. There are eight
memory card slots available for pSeries 690 and four for pSeries 670. For
detailed information about memory, see 2.3.2, “Memory subsystem for pSeries
690” on page 29, and 2.3.4, “Memory subsystem for pSeries 670” on page 42.
Memory slot 0
Memory slot 1
Memory slot 7
Memory slot 4
Memory slot 3
Memory slot 2
Memory slot 6
Memory slot 5
Figure 2-2 CEC front view
The rear view of CEC is shown in Figure 2-3 on page 21. The following slots are
provided:
򐂰 Slots for distributed converter assembly (DCA) books and two capacitor
books
These slots are populated by up to six DCA books and up to two capacitor
books. They supply electricity power to the CEC backplane and convert
voltage.
20
IBM ^ pSeries 670 and pSeries 690 System Handbook
򐂰 GX bus slots 0-3
The GX bus slot 0 is used to insert the primary I/O book. The GX bus slots 1,
2, and 3 are used for the optional secondary I/O books. In Figure 2-3, the slot
2 is populated with an IO book, while slots 1 and 3 are not. The pSeries 670 is
restricted to at most one secondary I/O book, while the pSeries 690 can have
from zero to three secondary books. For detailed information about the order
in which to populate the GX slots, see 2.3.3, “MCMs and GX slots relationship
for pSeries 690” on page 39, and for description of the books, see 2.3.6, “I/O
books” on page 47.
DCA books
Secondary
Reserved
I/O
book 1
(GX Bus Slot 1)
Primary I/O book
(GX Bus Slot 0)
Secondary
Secondary
I/O book 2
(GX Bus Slot 2)
I/O book 3
(GX Bus Slot 3)
Capacitor books
Figure 2-3 CEC rear view
In the center of the CEC, a CEC backplane is positioned vertically. As shown in
Figure 2-4 on page 22, it provides mount spaces for up to four multichip modules
(MCMs), sixteen level 3 (L3) cache modules, the eight memory cards and the
four I/O books. In the center position of the backplane is the clock card. It
distributes sixteen paired clock signals to the four MCMs. Please note that the
four GX bus slots (shown as “I/O Card Slots Books” in Figure 2-4 on page 22) are
located in the rear side of the CEC backplane.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
21
Figure 2-4 CEC backplane orthogonal view
2.3.1 POWER4 processor and MCM packaging
Several years ago, scientists working for microprocessor development set out to
design a new microprocessor that would leverage IBM strengths across many
different disciplines to deliver a server that would redefine what was meant by the
term server. POWER4 is the result. It was developed by over 300 engineers in
several IBM development laboratories. A new generation of this chip is available
since May 2003, which is called POWER4+. It has the same architecture as the
initial POWER4 chip, but is able to sustain higher clock frequency.
The pSeries 670 and pSeries 690 are based on these high performance
POWER4 chips. Up to 32 processors can be configured, and they are mounted
together in multichip modules. This packaging methodology was carefully
designed to provide increased levels of reliability by eliminating the tiered
packaging levels of separate processor modules mounted on processor cards;
these were mounted on a backplane. A pSeries 690 can have up to four MCMs
installed.
The pSeries 670 is using the same high performance POWER4 chips, but it only
supports up to 16 processors, which is a total of two MCMs installed.
Several versions of the POWER4 chips are available on pSeries 670 and pSeries
690 servers, with either 4 or 8 CPUs, executing at different speeds: 1.1GHz, 1.3
22
IBM ^ pSeries 670 and pSeries 690 System Handbook
GHz, 1.5 GHz or 1.7 GHz. All processors within pSeries 670 or pSeries 690
server, must be identical and operate at the same speed. The following table
describes the supported combinations of processors.
Table 2-2 Supported combinations of processors
Architecture
Clock
frequency
1.1 GHz
POWER4
1.3 GHz
1.5 GHz
Number of
Processors
per MCM
pSeries 670
pSeries 690
Designation
4
4-way
N/A
Standard
8
8-way or
16-way
8-way, 16-way,
24-way or
32-way
Standard
4
N/A
8-way or
16-way
HPC
8
N/A
8-way, 16-way,
24-way or
32-way
Turbo
4
4-way
N/A
8
8-way or
16-way
8-way, 16-way,
24-way or
32-way
8
N/A
8-way, 16-way,
24-way or
32-way
POWER4+
1.7 GHz
Turbo
The 1.1 GHz and 1.3 GHz 4-way MCMs are slightly different from the others.
They use a POWER4 chip with a single core. The 1.3 GHz 4-way MCM (called
the HPC feature) is specifically described in “Single core POWER4 processor
feature” on page 28. It is a special option for technical applications that demand
high memory bandwidth.
The chip has two processor cores and a level 2 (L2) cache, all in the same silicon
substrate. The L2 cache is shared between the two cores through a crossbar
switch known as the core interface unit (CIU). All these components are part of a
single POWER4 chip, along with communication buses and a level 3 (L3) cache
directory.
The following sections give a brief explanation about the POWER4 and MCM
technologies.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
23
For further detailed information about POWER4 and memory subsystem in the
pSeries 690, refer to the following whitepapers and redbook:
򐂰 POWER4 System Microarchitecture, found at:
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/power4.html
򐂰 IBM ^ pSeries 690 Configuring for Performance, found at:
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_config.html
򐂰 The POWER4 Processor Introduction and Tuning Guide, SG24-7041
POWER4 core
The internal micro-architecture of the POWER4 core is a speculative superscalar
Out-of-Order execution design. Up to eight instructions can be issued each cycle,
with a sustained completion rate of five instructions. Register rename pools and
other Out-of-Order resources coupled with the pipeline structure allow the design
to have over 200 instructions in flight at any given time.
In order to exploit instruction level parallelism, there are eight execution units,
each capable of being issued an instruction each cycle. Two identical
floating-point execution units, each capable of starting a fused multiply and add
each cycle are provided. This combination can achieve four floating-point
operations (FLOPs) per cycle per core. In order to feed the dual floating-point
units, two load/store units, each capable of performing address generation
arithmetic, are provided.
The microprocessor cores inside a POWER4 chip are a new design, although
they support the same PowerPC® instruction set architecture as prior pSeries
processors. Each core has several execution units, along with level 1 (L1)
instruction and data caches, and run at clock speeds of up to 1.7 GHz. Each L1
instruction cache is 64 KB and each L1 data cache is 32 KB. All data stored in
the L1 data cache is available in the L2 cache, guaranteeing no data loss.
The initial POWER4 manufacturing process, known as CMOS-8S3SOI,
implements seven-layer copper metallization, and 0.18 µm SOI CMOS
technology. The POWER4+ manufacturing process is called CMOS-9SSOI with a
0.13 µm SOI CMOS technology. Along with performance improvements, these
technologies deliver high-reliability components that IBM considers to be
fundamental for high-end, continuous operation servers. All the buses scale with
processor speed.
24
IBM ^ pSeries 670 and pSeries 690 System Handbook
L2 cache subsystem
The L2 cache on the POWER4 chip is composed of three separate cache
controllers, connected to the two processor cores through a core interface unit. In
POWER4 and POWER4+ chips, each L2 cache contains, respectively, 480 KB
and 512KB, for a total of 1.44 MB per POWER4 chip, and 1.5 MB per POWER4+
chip.
Each L2 cache controller can operate concurrently and feed 32 bytes of data per
cycle. The CIU connects each of the three L2 controllers to either the L1 data
cache or the L1 instruction cache in either of the two processors.
Additionally, the CIU accepts stores from the processors across 8-byte wide
buses and sequences them to the L2 controllers. Each processor has a
non-cacheable (NC) unit associated with it, responsible for handling instruction
serializing functions and performing any non-cacheable operations in the
memory hierarchy. In a logical view, the NC units are part of the L2 cache.
The L2 cache on the POWER4 chip is dedicated to delivering data to the two
cores as fast as they can process it, maximizing the overall performance. It has
multiple ports and is capable of multiple concurrent operations. The L2 cache
can provide data at the following peak rates:
򐂰
򐂰
򐂰
򐂰
򐂰
1.1 GHz Standard system: 105.6 GB/s to the two cores
1.3 GHz Turbo feature: 124.8 GB/s shared to the two cores
1.3 GHz HPC feature: 83.2 GB/s to the single core
1.5 GHz POWER4+ system: 144 GB/s shared to the two cores
1.7 GHz POWER4+Turbo feature: 163.2 GB/s shared to the two cores
L3 cache controller and directory
Each POWER4 processor chip, not each core, controls one L3 cache module.
The L3 module itself is located outside the chip, but the L3 directory and L3
controller are located on the POWER4 chip. A separate functional unit, referred
to as the Fabric Controller, is responsible for controlling data flow between the L2
and L3 controller for the chip and for POWER4 communication.
Each POWER4 chip has its own interface to the off chip L3 across two 16-byte
wide buses, operating at one third of the processor frequency. To communicate
with I/O devices, two 4-byte wide GX buses, operating at one third processor
frequency, are used. Finally, each chip has its own Joint Test Action Group
(JTAG) interface to the system service processor.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
25
Also included on the chip are functions called pervasive functions. These include
trace and debug facilities used for First Failure Data Capture (FFDC), Built-in Self
Test (BIST) facilities, Performance Monitoring Unit, an interface to the service
processor used to control the overall system, Power On Reset (POR)
Sequencing logic, and error detection and logging circuitry.
Multichip modules
The POWER4 chips are packaged on a single module called multichip modules.
Each MCM houses four chips (eight CPU cores) that are connected through
chip-to-chip ports.
The chips are mounted on the MCM such that they are all rotated 90 degrees
from one another, as shown in Figure 2-5. This arrangement minimizes the
interconnect distances, which improves the speed of the inter-chip
communication. There are separate communication buses between processors
in the same MCM, and processors in different MCMs (see Figure 2-6 on
page 27).
8-processor Multi-Chip Module
Chips rotated to allow faster
interconnection
Figure 2-5 POWER4 multichip module
26
IBM ^ pSeries 670 and pSeries 690 System Handbook
4.5" square, glass-ceramic
construct
An internal representation of the MCM is shown in Figure 2-6, with four
interconnected POWER4 chips. Each installed MCM comes with 128 MB of L3
cache. This provides 32 MB of L3 cache per POWER4 chip. The system bus (L3
cache, GX Bus, memory nest) operates at a 3:1 ratio with the processor
frequency. Therefore the L3 cache to MCM connections operate at:
򐂰
򐂰
򐂰
򐂰
375 MHz for 1.1 GHz processors
433 MHz for 1.3 GHz processors
500 MHz for 1.5 GHz processors
567 MHz for 1.7 GHz processors
4 POWER4 chips (8 processors) on an MCM
>1 Ghz
Core
L3
Shared L2
>1 Ghz
Core
Shared L2
L3 Dir
M
E
M
O
R
Y
Mem
Ctrl
L3
Chip-chip communication
GX Bus
Mem
Ctrl
GX Bus
4 GX links for
external
connections
GX Bus
Mem
Ctrl
GX Bus
L3
Shared L2
Shared L2
L3
Mem
Ctrl
M
E
M
O
R
Y
Multi-chip Module
L3 cache shared across all processors
(If the amount of memory behind each L3 is the same size.)
Figure 2-6 Multichip module with L2, L3, and memory
The MCM is a proven technology that IBM has been using for many years in the
mainframe systems (now IBM ^ zSeries). It offers several benefits in
mechanical design, manufacturing, and component reliability. IBM has also used
MCM technology in the RS/6000 servers in the past. The IBM RS/6000 Model
580 was based on the POWER2™ processor that has all its processing units and
chip-to-chip wiring packaged in an MCM.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
27
Single core POWER4 processor feature
As stated before, some technical applications benefit from very large bandwidth
between processors and memory. The POWER4 processor delivers an
exceptional bandwidth to the cores inside. For those applications that require
extremely high bandwidth, the high performance computing (HPC) feature is an
attractive alternative. Instead of 8-way MCMs, you have 4-way MCMs with the
same amount of L2 and L3 caches and the same bus interconnection (see
Figure 2-7).
This configuration provides twice the amount of L2 and L3 cache per processor
and additional memory bandwidth, when compared to the pSeries 690
configured with 8-way processor MCMs. This additional cache and memory
bandwidth available for each processor in this configuration may provide
significantly higher performance per processor for certain engineering and
technical environment applications.
Note: The HPC feature is only available on the pSeries 690 with a POWER4
1.3 GHz processor. It is not offered with the POWER4+ 1.5 or 1.7 GHz
processors.
28
IBM ^ pSeries 670 and pSeries 690 System Handbook
pSeries 690 1.3 GHz Turbo
Dual CPU Core
D
1.3 GHz
CPU
I
1.3 GHz
CPU
D
I
Shared L2 Cache
Cache Architecture
L1 - Data: 32 KB Instruction: 64 KB
L3 Controller
L3 Directory
Fabric Controller
Distributed Switch
L2 - Shared ~1440 KB @ 124.8 GB/sec
L3 - Shared 32 MB/chip (128 MB/8-way module) @ 13.9 GB/sec L3 Cache
Interprocessor bus
I/O Bus
Memory
pSeries 690 1.3 GHz HPC
D
I
1.3 GHz
CPU
Single D
>1GHz CPU
CPU Core I
Dedicated L2 Cache
L3 Controller
L3 Directory
Cache Architecture
L1
Data: 32 KB Instruction: 64 KB
L2
Dedicated 1440 KB @ 83.2 GB/sec
L3
32 MB/chip (128MB/4-way module) @ 13.9 GB/sec
L3 Cache
Fabric Controller
Distributed Switch
Interprocessor bus
I/O Bus
Memory
Figure 2-7 pSeries 690 1.3 GHz Turbo and 1.3 GHz HPC feature
2.3.2 Memory subsystem for pSeries 690
The memory subsystem in the pSeries 690 is physically composed of L3 cache
modules and up to eight memory cards, both mounted on the CEC backplane.
Since each L3 cache module is controlled by an L3 controller that physically
resides in a POWER4 processor chip packaged in an MCM, there are tight
relationships between MCMs, L3 cache modules, and memory cards.
MCM population order
The pSeries 690 can be configured with up to four MCMs mounted on the
backplane, as shown in Figure 2-8. The MCM 0 is always configured. The other
optional MCMs are populated in this order: MCM 2, MCM 1, and MCM 3. In a
two-MCM configuration (MCM 0 and MCM 2), because the spaces for MCM 1
and MCM 3 are not populated, you have to have a two-processor bus
pass-through modules in this space to establish the link between two populated
MCMs. In a three-MCM configuration (MCM 0, MCM 1, and MCM 2), since the
space for MCM 3 is not populated, you have to have a processor bus pass
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
29
through a module in this space. For the configuration rules of multiple MCMs, see
3.2.3, “Processor configuration rules” on page 85.
Figure 2-8 MCM, L3 cache, and memory slots relationship on backplane
Figure 2-9 provides a logical view of the relationship between MCMs and
memory slots.
30
IBM ^ pSeries 670 and pSeries 690 System Handbook
Mem
Slot 6
GX
Slot 3
GX
Mem Slot 0
Slot 7
Mem
Slot 0
L3 L3
L3 L3
GX
L3 L3
GX
L3 L3
GX
L3 L3
GX
L3 L3
GX
Mem
Slot 2
GX
GX
P P
P P
P P
L2
L2
L2
L2
P P
P P
P P
P P
L2
L2
L2
L2
P P
P P
P P
P P
L2
L2
L2
L2
MCM 3
MCM 2
L2
L3 L3
P P
P P
L2
L2
L2
GX
Slot 2
L3 L3
GX
GX
GX
L3 L3
L3 L3
GX
P P
L3 L3
L3 L3
GX
MCM 1
GX
L3 L3
Mem
Slot 1
GX
Slot 1
GX
MCM 0
GX
Mem
Slot 3
L3 L3
P P
P P
GX
L3 L3
L3 L3
Mem
Slot 5
Mem
Slot 4
Figure 2-9 Logical relationship between MCMs, memory, and GX Slots
Multiple MCM configuration
In a four-MCM configuration, there is no processor bus pass-through module
required, because all the space is populated with MCMs. In this configuration,
there are sixteen 32 MB L3 caches, 512 MB total, and four rings interconnecting
the MCMs, as shown in Figure 2-10 on page 32.
In two- and three-MCM configurations, unpopulated spaces are replaced with
processor bus pass-through modules. If one of the MCMs is replaced with a
processor bus pass-through module, MCMs are still connected through a
four-ring interconnection. However you lose access to the specific memory slots
under the unpopulated MCMs, because a processor bus pass-through module
does not provide access to the memory slots directly connected to it. The
processor bus modules also do not have access to the GX bus (see 2.3.3,
“MCMs and GX slots relationship for pSeries 690” on page 39).
Forming the four rings interconnecting the MCMs, a processor can access all the
memory cards.
򐂰 If the referenced address is in the memory cards under the directly attached
L3 cache to the MCM, then interconnection between processor cores within
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
31
MCM is used to reference the memory. This access is explained in “Single
MCM configuration” on page 32.
򐂰 If the referenced address is not in the memory cards under the directly
attached L3 cache to the MCM, then interconnection between MCMs is used
for access. This memory access is achieved by a very small overhead.
See Figure 2-10 for more details about the interconnection between the four
MCMs.
MCM 3
MCM 0
MCM 2
MCM 1
Figure 2-10 Interconnection between four MCMs
Single MCM configuration
In a single MCM configuration (always MCM 0), there is no processor buss
pass-through module required, because there are no interconnections between
MCMs. In this configuration, a processor can access the memory cards in slot 0
and 1, but cannot access the other memory cards. Within this MCM, each
processor can access all the memory addresses through its own L3 cache, or
through the interconnection between processors within MCM, as shown in
Figure 2-11.
32
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 2-11 Interconnection between processors in an MCM
Access to memory through L3 cache
Each POWER4 processor chip controls a 32 MB L3 cache, and each MCM has
access to two memory cards through these four L3 caches, 128 MB total, as
shown in Figure 2-11.
The general rule is that the two memory slots attached to an MCM are populated
with two identically sized memory cards. Then these four L3 caches act as a
single interleaved 128 MB sized L3 cache. Combining L3 cache chips into a
group not only increases the L3 cache size, but also scales the available L3
bandwidth. When combined into a group, four L3 caches and the memory behind
them are interleaved on 512 byte granularity.
In two special cases (Single 8 GB memory card attached to an 8-way system and
single 4 GB memory card in a pSeries 670 system), the two memory slots are
populated with two differently sized memory cards: one of the memory slots is
unpopulated. Then the four L3 caches act as two separate 64 MB sized L3
caches. For optimal performance, memory cards of equal capacity should be
attached to each MCM.
Each memory card has two memory controller chips, which are connected to an
L3 cache module and to synchronous memory interface (SMI) chips.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
33
Each memory controller chip can have one1 or two2 memory data ports and can
support up to 32 GB of memory.
Two sets of memory cards can be found in the pSeries 690 servers:
򐂰 The cards in the first set sustain a bus frequency up to 500 MHz. They are
exist in several sizes: 4 GB, 8 GB, 16 GB, and 32 GB (the 16 GB, and 32 GB
cards will no longer be available after May 2003).
򐂰 The cards in the second set sustain a bus frequency up to 567 MHz. They are
available in several sizes: 4 GB, 8 GB, 16 GB, 32 GB, and 64 GB.
All models of pSeries 670 and pSeries 690 can use the 500 MHz cards and the
16 GB, 32 GB, and 64 GB at 567 MHz cards. The 4GB and 8GB at 567 MHz
memory features are only available with the 1.7 GHz processor configurations.
The memory controller is attached to the L3 modules, with each module having
two 16-byte buses to the data interface in the memory controller. These buses
operate at one-third processor speed using the Synchronous Wave Pipeline
Interface to operate at high frequencies.
Restriction: If a 500 MHz card is used with a 1.7 GHz processor, it is not able
to run at a 3:1 ratio of the processor speed. In this case, it will be detected as
a slow card, and as a result, all memory buses and all memory cards in the
system will operate at a 4:1 ratio of the processor speed: 425 MHz.
The memory modules are Dual Data Rate (DDR) and Synchronous DRAM
(SDRAM), soldered on the memory card for improved reliability, and use book
packaging.
Memory card configuration for pSeries 690
The pSeries 690 has eight memory slots. Memory is available in two types:
Inward-facing and outward-facing. The inward-facing memory slots (0, 1, 2, and
3) are enabled by the MCM 0 and MCM 2, while the outward-facing memory slots
(4, 5, 6, and 7) are enabled by the MCM 1 and MCM 3 (see Figure 2-8 on
page 30).
As you add more MCMs, you activate more memory ports. It increases the
configurable memory size, but also increases the memory bandwidth. Table 2-3
summarizes this relationship.
1
2
In 4 GB and 8 GB memory cards.
In 16 GB, 32 GB, and 64 GB memory cards.
34
IBM ^ pSeries 670 and pSeries 690 System Handbook
Table 2-3 Relationship between MCMs, L3 cache, memory slots, and size
Populated MCMs
L3 cache modules
Usable memory
slots
Maximum
memory size
1
4
2
128 GB
2
8
4
256 GB
3
12
6
384 GB
4
16
8
512 GB
The minimum sized memory, 8 GB, is available with two 4 GB memory cards on
a single MCM or with a single 8 GB memory card. The maximum sized memory,
512 GB, is available in a four-MCM configuration, which supports a full eight
memory slots, with each populated by a 64 GB memory card.
Note: The whole memory related to an MCM is available to the system, even if
all the processors on this MCM are not activated. Customers with applications
requiring very large amounts of memory can therefore use the CUoD feature
to increase the maximum memory size supported on their server. For
example, a server with one 8-way MCM can have up to 128 GB of memory,
while an 8-way server with 2 CUoD MCMs (4 active processors and 4 inactive
processors) supports up to 256 GB of memory. See Chapter 4, “Capacity
Upgrade on Demand” on page 133 for details on memory configurations
supported with the CUoD feature.
The optimal amount of memory for a particular system depends upon many
factors, such as the requirements of the key applications targeted for the system.
However, it is the size and number of memory cards in the system that determine
the maximum bandwidth the system can deliver.
Memory should be installed and balanced across all populated MCM positions
for optimal performance. Unbalanced configurations will function properly, but the
unbalanced portion of the memory will not utilize the full memory bus bandwidth.
Table 2-4 provides supported memory card placement for various memory
requirements. The memory configurations with (*) are valid but not
recommended. You have to review application performance implications before
ordering this memory configuration.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
35
Table 2-4 Supported memory cards configurations
Number of
MCMs
Memory
size in GB
Memory
slots 0 & 1
Memory
slots 2 & 3
8
4+4
8 (*)
8+0
16
8+8
32
16 + 16
64
32 + 32
128
64+64
8
4+4
0+0
16
4+4
4+4
16 (*)
8+8
0+0
24
8+8
4+4
32
8+8
8+8
48
16 + 16
8+8
64
16 + 16
16 + 16
96
32 + 32
16 + 16
128
32 + 32
32 + 32
192
64 + 64
32 + 32
256
64 + 64
64 + 64
1
2
36
IBM ^ pSeries 670 and pSeries 690 System Handbook
Memory
slots 4 & 5
Memory
slots 6 & 7
Number of
MCMs
Memory
size in GB
Memory
slots 0 & 1
Memory
slots 2 & 3
Memory
slots 4 & 5
8
4+4
0+0
0+0
16
4+4
4+4
0+0
16 (*)
8+8
0+0
0+0
24
4+4
4+4
4+4
24 (*)
8+8
4+4
0+0
32
8+8
4+4
4+4
32 (*)
8+8
8+8
0+0
40
8+8
8+8
4+4
48
8+8
8+8
8+8
64
16 + 16
8+8
8+8
80
16 + 16
16 + 16
8+8
96
16 + 16
16 + 16
16 + 16
128
32 + 32
16 + 16
16 + 16
160
32 + 32
32 + 32
16 + 16
192
32 + 32
32 + 32
32 + 32
256
64 + 64
32 + 32
32 + 32
320
64 + 64
64 + 64
32 + 32
384
64 + 64
64 + 64
64 + 64
Memory
slots 6 & 7
3
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
37
Number of
MCMs
Memory
size in GB
Memory
slots 0 & 1
Memory
slots 2 & 3
Memory
slots 4 & 5
Memory
slots 6 & 7
8
4+4
0+0
0+0
0+0
16
4+4
4+4
0+0
0+0
16 (*)
8+8
0+0
0+0
0+0
24
4+4
4+4
4+4
0+0
24 (*)
8+8
4+4
0+0
0+0
32
4+4
4+4
4+4
4+4
32 (*)
8+8
4+4
4+4
0+0
32 (*)
8+8
8+8
0+0
0+0
40
8+8
4+4
4+4
4+4
40 (*)
8+8
8+8
4+4
0+0
48
8+8
8+8
4+4
4+4
48 (*)
8+8
8+8
8+8
0+0
56
8+8
8+8
8+8
4+4
64
8+8
8+8
8+8
8+8
80
16 + 16
8+8
8+8
8+8
96
16 + 16
16 + 16
8+8
8+8
112
16 + 16
16 + 16
16 + 16
8+8
128
16 + 16
16 + 16
16 + 16
16 + 16
160
32 + 32
16 + 16
16 + 16
16 + 16
192
32 + 32
32 + 32
16 + 16
16 + 16
224
32 + 32
32 + 32
32 + 32
16 + 16
256
32 + 32
32 + 32
32 + 32
32 + 32
320
64 + 64
32 + 32
32 + 32
32 + 32
384
64 + 64
64 + 64
32 + 32
32 + 32
448
64 + 64
64 + 64
64 + 64
32 + 32
512
64 + 64
64 + 64
64 + 64
64 + 64
4
38
IBM ^ pSeries 670 and pSeries 690 System Handbook
In the configurations listed previously in Table 2-4, you can notice that the first
memory slots are populated with the largest memory boards in the first memory
slots. This corresponds to the recommended order for inserting memory boards,
starting with the largest boards in the first slots, and then installing the following
boards in decreasing memory size order. IBM manufacturing will deliver the initial
configuration of pSeries 690 with memory installed according to this rule.
However, this order is only a recommendation, and there may be cases where it
is not respected. For example, if a customer has a system initially configured with
48 GB memory as follows:
16 + 16
8+8
Then the customer orders an MES to increase the memory to 96 GB with one
pair of 16 GB memory board and one pair of 8 GB memory board (assuming the
system is populated with 4 MCM), the memory configuration will become:
16 + 16
8+8
16 + 16
8+8
This is a valid and supported configuration. It would not be possible to reorder
the memory boards since the 8 GB memory boards initially installed in slots 2
and 3 are inward facings and could not be moved to slots 4, 5, 6, or 7 when the
MES memory boards are installed.
Note: Consideration for system upgrades. When initially ordering a pSeries
690 configuration with the intent to order additional hardware later, you must
pay attention to the supported memory configurations available with different
numbers of MCM. For example, if you order a two MCM pSeries 690 with 256
GB of memory, and later you want to order two extra MCM, you will have to
also order at least 128 GB extra memory.
For more information on physical memory configuration and performance
considerations, please refer the IBM ^ pSeries 690 Configuring for
Performance white paper, found at:
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_config.html
2.3.3 MCMs and GX slots relationship for pSeries 690
The number of installed MCMs also affects the number of I/O drawers supported.
An MCM is connected to two GX slots through two GX buses using four ports.
The GX slot3 is a physical connector that is used for inserting an I/O book. For
detailed information about I/O books, see 2.3.6, “I/O books” on page 47.
3
Although the GX bus and the GX slot are physically different components, from here on we use GX slot to express both
of them, since there is a one-to-one relationship between a GX bus and a GX slot.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
39
If there is direct connection to the specific GX slot between a POWER4 chip and
this slot, then this path is used. If the GX slot is being accessed from the other
chip within an MCM, then an inter-chip connection link is used. If this GX slot is
being accessed from a processor chip in the other MCMs only, then an
inter-MCM connection link is used. Figure 2-9 on page 31 provides a logical view
of the relationship between MCMs and GX slots.
As explained in 2.3.6, “I/O books” on page 47, the primary I/O book has four
Remote I/O (RIO) ports, and each of the three optional secondary I/O books has
eight RIO ports each.
The GX slots can be populated with IO books in a predefined order: GX Slot 0,
then GX Slot 2, GX Slot 3 and finally GX Slot 1. The primary IO book in
GX Slot 0 is always installed. The number of installed MCM defines how many
secondary IO books can be installed in the GX slots:
򐂰 In a single MCM configuration (only MCM 0 is populated), the POWER4
processor chips 0 and 2 are connected to GX Slot 0, and the processor chips
1 and 3 are connected to GX Slot 1, as shown in Figure 2-12 on page 41. But
since the GX Slot 1 cannot be used in a single MCM configuration, only the
primary I/O book is accessible in GX Slot 0. This means that up to two I/O
loops can be configured.
򐂰 In a two-MCM configuration, GX Slots 0, 2 and 3 can be populated with IO
books. Therefore, 20 RIO ports are physically available, shown in the column
Physically Available RIO ports in Table 2-5. Actually, in a two-MCM
configuration (MCM 0 and MCM 2 are populated), GX Slot 2 and 3 are
accessed from the MCM 2 only. Therefore, in addition to the four ports on GX
slot 0, only four out of eight IO ports on GX Slot 2, and only four RIO ports out
of eight RIO ports on GX Slot 3 are accessed, for a total of 12 usable ports:
only 6 I/O loops4 are supported at most in a two-MCM configuration.
򐂰 In a three-MCM configuration, the four GX bus can be populated with RIO
books. Only four RIO ports on GX Slot 3 can be used. A total of 24 RIO ports
are usable, for a maximum of 12 RIO loops.
򐂰 In a four-MCM configuration, all 28 RIO ports can be used, providing
connectivity for up to 14 I/O loops.
4
An I/O loop requires two RIO ports to be connected.
40
IBM ^ pSeries 670 and pSeries 690 System Handbook
MCM 0
Primary I/O book
Processor
chip 0
GX-RIO 0
Processor
chip 2
GX-RIO 1
Processor
chip 3
Processor
chip 1
RIO-PCI
Service
Processor
GX slot 0
HMC ports
OP-Panel
Loops to RIO
drawers
Diskette
MCM 1
Secondary I/O book 1
Processor
chip 4
GX-RIO 2
Processor
chip 6
GX-RIO 3
Processor
chip 7
GX-RIO 8
Processor
chip 5
GX-RIO 9
GX slot 1
MCM 2
Secondary I/O book 2
Processor
chip 8
GX-RIO 10
Processor
chip 10
GX-RIO 11
Processor
chip 11
GX-RIO 4
Processor
chip 9
GX-RIO 5
GX slot 2
MCM 3
Secondary I/O book 3
Processor
chip 12
GX-RIO 6
Processor
chip 14
GX-RIO 7
Processor
chip 15
GX-RIO 12
Processor
chip 13
GX-RIO 13
GX slot 3
Figure 2-12 MCMs and RIO ports relationship (pSeries 690)
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
41
We summarize the supported RIO ports numbers in Table 2-5.
Table 2-5 Relationship between MCMs, RIO ports, and I/O drawers
Populated MCMs
Available GX
Slots
Number of
Physically
existing RIO ports
Number of usable
RIO ports
MCM 0
0
4
4
MCM 0, 2
0, 2, 3
20
12
MCM 0, 1, 2
0, 1, 2, 3
28
24
MCM 0, 1, 2, 3
0, 1, 2, 3
28
28
Note: As for memory, the limitations on the number of supported GX slots and
IO ports depends on the number of installed MCM, and not the number of
activated processors. Therefore, the CUoD feature can be used when a large
number of IO drawers are required on a system with “few” processors. For
example, a server with one 8-way MCM can have up to two I/O drawers, while
a 8-way server with 2 CUoD MCMs (4 active processors and 4 inactive
processors) supports up four 4 I/O drawers.
2.3.4 Memory subsystem for pSeries 670
This section only mentions information related to the memory subsystem, which
are different for pSeries 670 and for pSeries 690.
Information presented in the these three sections for the pSeries 690 also apply
to the pSeries 670:
򐂰 “MCM population order” on page 29
򐂰 “Multiple MCM configuration” on page 31
򐂰 “Single MCM configuration” on page 32
Besides the fact that only the number of MCMs is restricted to two for pSeries
670, the backplane is basically the same as for the pSeries 690. Only two MCMs
and four memory slots can be used. MCM positions MCM1 and MCM3 are
unused. The minimum required memory with one MCM installed is 4 GB. The
maximum memory available with two MCMs is 256 GB. Only the inner memory
slots can be used and should be populated in identical pairs. Figure 2-13 on
page 43 shows the backplane of pSeries 670.
42
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 2-13 MCM, L3 cache, and memory slots relationship on backplane
Memory cards configuration for pSeries 670
The first MCM activates the first and second memory positions. The second
MCM activates the remaining third and fourth positions. For best performance
you should populate all activated memory card positions with memory cards, and
memory sizes should be balanced as closely as possible across all populated
MCM locations.
As for pSeries 690, two sets of memory cards can be found in the pSeries 670
servers:
򐂰 The 500 MHz set, with memory boards available in sizes: 4 GB, 8 GB, 16 GB,
and 32 GB.
򐂰 The 567 MHz set, with memory boards available in sizes: 16 GB, 32 GB, and
64 GB.
The memory boards operate at a 1:3 ratio of the processor speed. Table 2-6
shows the supported memory card configurations on pSeries 670.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
43
Table 2-6 Supported memory configurations for pSeries 670
MCMs
installed
1
2
Memory size in GB
Memory slots 0 & 1
Memory slots 2 & 3
4
4+0
8
4+4
8a
8+0
16
8+8
32
16 + 16
64
32 + 32
128
64+64
8
4+4
0+0
16
4+4
4+4
16a
8+8
0+0
24
8+8
4+4
32
8+8
8+8
48
16 + 16
8+8
64
16 + 16
16 + 16
96
32 + 32
16 + 16
128
32 + 32
32 +32
192
64+64
32 + 32
256
64+64
64+64
N/A
a. This is a valid configuration, but not a recommended configuration.
2.3.5 MCMs and GX slots relationship for pSeries 670
The pSeries 670 shares the same MCM-to-GX slot relationship with pSeries 690,
although it only supports up to two MCMs and a maximum of three I/O drawers.
Figure 2-14 on page 46 illustrates the relationship between MCMs and GX slots
on the pSeries 670. Only two I/O books (GX slot 0 and 2) are supported.
The number of installed MCM defines how many secondary IO books can be
installed in the GX slots.
44
IBM ^ pSeries 670 and pSeries 690 System Handbook
򐂰 In a single MCM configuration (only MCM 0 is populated), the POWER4
processor chips 0 and 2 are connected to GX slot 0, and the processor chips
1 and 3 are connected to GX slot 1, as shown in Figure 2-12 on page 41. But
since the GX slot 1 cannot be used in a pSeries 670 configuration, only the
primary I/O book is accessible in GX slot 0. This means that up to two I/O
loops can be configured.
򐂰 In a two-MCM configuration, GX slots 0 and 2 can be populated with IO
books. Therefore, 12 RIO ports are physically available, shown in the column
Physically Available RIO ports in Table 2-5. Actually, in a two-MCM
configuration (MCM 0 and MCM 2 are populated), GX slot 2 is accessed from
the MCM 2 only. Therefore, in addition to the four ports on GX slot 0, only four
out of eight IO ports on GX slot 2 are accessed, for a total of 8 usable ports:
only 4 I/O loops5 are supported at most in a two-MCM configuration.
Table 2-7 summarizes the supported RIO ports numbers.
Table 2-7 Relationship between MCMs, RIO ports, and I/O drawers
5
Populated MCMs
Available GX
Slots
Number of
Physically
existing RIO ports
Number of usable
RIO ports
MCM 0
0
4
4
MCM 0, 2
0, 2
12
8
An I/O loop requires two RIO ports to be connected.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
45
MCM 0
Primary I/O book
Processor
chip 0
GX-RIO 0
Processor
chip 2
GX-RIO 1
Processor
chip 3
Processor
chip 1
RIO-PCI
Service
Processor
GX slot 0
HMC ports
OP-Panel
Diskette
MCM 1
Processor Bus
Path Through
Module
Reserved
GX slot 1
MCM 2
Secondary I/O book
Processor
chip 8
GX-RIO 10
Processor
chip 10
GX-RIO 11
Processor
chip 11
GX-RIO 4
Processor
chip 9
GX-RIO 5
GX slot 2
MCM 3
Processor Bus
Path Through
Module
Reserved
GX slot 3
Figure 2-14 MCMs and RIO ports relationship (pSeries 670)
46
IBM ^ pSeries 670 and pSeries 690 System Handbook
Loops to RIO drawers
2.3.6 I/O books
The I/O books plug into the GX slots in the CEC backplane and provide the
Remote I/O (RIO) ports. The RIO ports are used to connect I/O drawers to the
GX bus. Each I/O book contains a base card, a riser, and a daughter card, and is
physically packaged using book packaging technology, explained in “Book
packaging” on page 49.
There are two types of I/O books available, called primary I/O book and
secondary I/O book and shown in Figure 2-15 on page 48. The primary I/O book
is mandatory in all pSeries 670 and pSeries 690 and there must be exactly one
per system. The secondary I/O books are optional. Their number depends on the
requirements for external IO drawer attachments. A pSeries 670 can contain at
most one secondary I/O book, while a pSeries 690 can contain up to three
secondary I/O books.
򐂰 The primary I/O book
– Contains the following ports:
•
Two HMC connections ports: HMC1 and HMC2
•
Two native serial ports: S1 and S2
These two native serial ports are under the control of the service
processor and are not available for functions, such as HACMP
heartbeat cabling or UPS control, which require fully dedicated ports.
•
An operator panel port
•
A diskette drive port
•
A bulk power controller (BPC) Y-cable connector
This port is used to connect between two BPCs on the bulk power
assembly (BPA) and the service processor using a Y-cable.
– Contains the service processor.
– Contains non-volatile random access memory (NVRAM).
– Contains four RIO ports for the first I/O drawer (mandatory) and an
optional second I/O drawer.
– This I/O book plugs into the GX slot 0 in the CEC backplane.
򐂰 Each secondary I/O book
– Contains eight RIO ports.
– Plugs into the GX slots of the CEC backplane in the order: GX slot 2, GX
slot 3 and GX slot 1. In Figure 2-15 on page 48, the RIO Port numbers
correspond to the first secondary I/O book. For the second and third
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
47
secondary books, the port number would be, respectively, 6, 7, 12, and 13,
and 2, 3, 8, and 9.
Primary I/O book
Secondary I/O book
RIO port 0 (A0)
Op Panel
RIO port 11 (B1)
RIO port
5 (D1)
RIO port 0 (A1)
RIO port 11 (B0)
S1
S2
RIO port
5 (D0)
RIO port 1 (B0)
HMC 1
RIO port 10 (A1)
RIO port
4 (C1)
HMC 2
RIO port 1 (B1)
Diskette
RIO port 10 (A0)
RIO port
4 (C0)
BPC Y-cable connector
Figure 2-15 Primary and secondary I/O books
For details on connections between the I/O books and the I/O drawers, see 2.4.2,
“I/O subsystem communication and monitoring” on page 56.
There are two sets of I/O books, which differ by the technology used in the base
card, a riser, and a daughter card installed in the book. The first set is referred to
as RIO while the second is called RIO-2.
򐂰 The RIO primary and secondary books are:
– Primary: FC 6404, Service Processor and RIO Loop Attachment, Two
Loops
– Secondary: FC 6410, RIO Loop Adapter, Four Loops
򐂰 The RIO-2 books are called:
– Primary: FC 6418, Service Processor and RIO-2 Loop Attachment, Two
Loops
– Secondary: FC 6419, RIO-2 Loop Adapter, Four Loops
48
IBM ^ pSeries 670 and pSeries 690 System Handbook
The RIO-2 books provide improvement in I/O performance and throughput
required by the new 1.5 and 1.7 GHz processors. The RIO-2 primary book
contains a larger NVRAM than the RIO primary book, and provide the increased
support from 16 to 32 LPARs. There are restrictions in the use of the books:
򐂰 All I/O books must be of the same technology. You cannot mix FC 6404 and
6419 or FC 6418 and FC6410 in the same system.
򐂰 The RIO books can only be used with the 1.1 and 1.3 GHz processors.
򐂰 The RIO-2 books can be used with the 1.1, 1.3, 1.5, and 1.7 GHz processors.
򐂰 When using RIO books, the maximum number of supported LPARs is 16.
򐂰 With RIO-2 books, it is possible to instantiate up to 32 LPARs, even with 1.1
GHz and 1.3 GHz systems.
Book packaging
Memory cards, I/O books, and capacitors in the CEC are packaged between
metal sheets to form what are called books. Book packaging helps protect
components from electrostatic discharge and physical damage, and also helps to
stabilize electronics and distribute air-flow throughout the CEC for proper
temperature control.
Frame guide rails, as shown in Figure 2-16 on page 50, help align the books
when connecting to the backplane; guidance pins assure final positioning and
two book locks secure the book to the backplane and frame cage. This results in
a reduction in pin damage when installing memory and I/O book upgrades, and
stabilizes the system for shipment and physical installation.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
49
BO O K
LO CK
FR A M E G U ID E R A IL
CONNECTOR
RIGID VERTICAL BACKPLANE
G UIDA NC E PIN
C O NNECTIN G
BO O K
TO B ACKPLA NE
G U ID A N CE PIN
FR A M E G UID E R A IL
Figure 2-16 Book packaging
2.3.7 Service processor
The service processor is a complete microcomputer, including random access
memory and program storage, within a computer system. This auxiliary
processor serves two main purposes:
򐂰 It initializes and tests the chip logic interconnects that constitute the processor
subsystem of the server, configuring them for normal operation.
򐂰 It constantly supervises the functional health of the server while the computer
system is in operation to detect any failing components as they occur.
The service processor is also responsible for hardware management and control,
and executes the required changes in the hardware configuration when creating
or modifying a partition. It is the interface between the pSeries 670 and pSeries
690 and the HMC. The service processor is packaged along with a quad RIO
port hub adapter in the primary I/O book.
The service processor is responsible for monitoring the entire system, and for
storing the machine configuration and partition information in NVRAM.
50
IBM ^ pSeries 670 and pSeries 690 System Handbook
2.4 I/O subsystem
The I/O drawer subsystem of the pSeries 670 and pSeries 690 uses similar
technology to that used in the IBM ^ pSeries 660 Model 6M1. The PCI
I/O adapters are housed in separate drawers that are connected to the CEC with
RIO cables.
Note: The pSeries 670 supports up to three I/O drawers, while the pSeries
690 supports up to eight I/O drawers. The pSeries 670 also supports a half I/O
drawer configuration (10 PCI slots).
For removable media, both pSeries 670 and pSeries 690 have a media drawer
with configurable devices that can be connected to different logical partitions.
2.4.1 I/O drawer
The I/O drawers provide internal storage and I/O connectivity to the system.
Figure 2-17 shows the rear view of an I/O drawer, with the PCI slots and riser
cards that connect to the RIO ports in the I/O books.
2 SCSI Backplanes
4 SCSI Drives each
2 SCSI Backplanes
4 SCSI Drives each
24 inch
10 PCI slots
RIO
Riser Card
Rear View
10 PCI slots
4U
RIO
Riser Card
Figure 2-17 I/O drawer rear view
Each drawer is composed of two physically symmetrical I/O planar boards that
contain 10 Hot-Plug PCI slots each, and PCI adapters can be inserted in the rear
of the I/O drawer. The planar boards also contain two integrated Ultra3 SCSI
adapters and SCSI Enclosure Services (SES), connected to a SCSI 4-pack
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
51
backplane. The pSeries 670 can be initially configured with only a half I/O drawer
with 10 PCI slots and up to eight disk drives installed.
As for I/O books, the I/O drawers exist in two technologies: RIO and RIO-2.
I/O drawers of both technologies offer the same number of PCI slots, the same
number of disks, and are packaged in the same chassis. The difference is the
type of I/O planar that is installed inside the drawer. Externally, the only visible
difference between the two technologies is the shape of the cable connectors on
the RIO Riser card (see Figure 2-18).
򐂰 RIO connectors have a thumbscrew retention physical connector.
򐂰 RIO-2 connectors have a bayonet retention physical connector.
RIO Connector
RIO-2 Connector
Figure 2-18 Difference between RIO and RIO-2 connectors
The I/O Drawer for pSeries 670 and pSeries 690 has its own product number:
7040-61D, which refers to the two technologies. For ordering purposes, the
difference between them is the Feature Code of the configured I/O planar:
򐂰 RIO drawer uses the “FC 6563, I/O Drawer PCI Planar, 10 slots, 2 Integrated
Ultra3 SCSI Ports”.
򐂰 RIO-2 drawer is configured with the “FC 6571, I/O Drawer PCI-X Planar, 10
slots, 2 Integrated Ultra3 SCSI Ports”.
A logical connection view of an RIO I/O drawer is shown in Figure 2-19.
52
IBM ^ pSeries 670 and pSeries 690 System Handbook
SCSI
4-pack
SCSI
4-pack
RIO Bus
I/O Controller
SCSI
4-pack
SCSI
4-pack
RIO Bus
I/O Controller
EADS
PCI- PCI
EADS
PCI- PCI
EADS
PCI- PCI
EADS
PCI- PCI
EADS
PCI- PCI
EADS
PCI- PCI
Bridge
Bridge
Bridge
Bridge
Bridge
Bridge
Ultra3
SES
SCSI
Ultra3
SES
SCS
Ultra3
SES
SCS
Ultra3
SES
SCSI
PCI Slot - 64 bit 66/33MHz 3.3V
PCI Slot - 64 bit 33MHz 5.0V
Figure 2-19 Logical view of an RIO drawer
For a RIO-2 I/O drawer, the topology of the logical connection view in identical, as
shown in Figure 2-20. The PCI-PCI bridge uses an EADS-X chip instead of an
EADS chip. The PCI slots’ differences are explained later in this section.
Link 0
Link 1
SCSI
4-pack
Link 0
Link 1 SCSI
4-pack
RIO2 to PCI-X
Bridge
SCSI
4-pack
RIO2 to PCI-X
Bridge
SCSI
4-pack
PHB1
PCI-X to PCI-X
PHB2
PCI-X to PCI-X
PHB3
PCI-X to PCI-X
PHB1
PCI-X to PCI-X
PHB2
PCI-X to PCI-X
Bridge
Bridge
PHB3
PCI-X to PCI-X
Bridge
Bridge
Bridge
Bridge
Planar 1
FC 6571
Ultra3
SES
SCSI
Ultra3
SCSI SES
Planar 2
FC 6571
Ultra3
SCSI SES
Ultra3
SCSI SES
PCI Slot - 64 bit 133MHz 3.3V
Figure 2-20 Logical view of an RIO -G drawer
Each of the 4-packs supports up to four hot-swappable Ultra3 SCSI disk drives,
which can be used for installation of the operating system or storing data.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
53
There are three different disk capacities: 18.2 GB, 36.4 GB, 73.4 GB and 146.8
GB, and they have the following characteristics:
򐂰 Form factor: 3.5-inch, 1-inch (25 mm) high
򐂰 SCSI interface: SCSI Ultra3 (fast 80) 16 bit
򐂰 Rotational speed: 10,000 RPM and 15,000 RPM
Optionally, external disk subsystems can be connected to serve as operating
system disks. The pSeries 670 and pSeries 690 and their logical partitions can
boot6 from SCSI, SSA, and Fibre Channel disk subsystems.
The RIO riser cards are connected to the planar boards. The RIO ports of each
riser card are connected through IO loops to RIO ports on I/O books in the CEC.
The connectivity between the I/O drawer RIO ports and the I/O books RIO ports
is described in “Remote I/O loop” on page 57.
The PCI slots have different characteristics in the RIO and RIO-2 drawers:
򐂰 RIO Drawer
On each planar board, the first seven PCI slots have a 3.3V PCI bus signaling
and operating at 66 MHz or 33 MHz, depending on the adapter. The last three
PCI slots have a 5V PCI bus signaling and operating at 33 MHz. All PCI slots
are PCI 2.2 compliant and are Hot-Plug enabled, which allows most PCI
adapters to be removed, added, or replaced without powering down the
system. This function enhances system availability and serviceability.
򐂰 RIO-2 Drawer
On each planar board, the ten PCI-X slots have a 3.3V PCI bus signaling and
operating at 33 MHz, 66MHz or 133 MHz, depending on the adapter. All PCI
slots are PCI 2.2 compliant and are Hot-Plug enabled.
PCI adapters have different bandwidth requirements, and there are functional
limitations on the number of adapters of a given type in an I/O drawer or a
system. The limitations differ for the RIO and RIO-2 drawers.
The limitations are of several types:
򐂰 Maximum number of adapters of one type per IO planar
򐂰 Maximum number of adapters of one type per IO drawer
򐂰 Maximum number of adapters of one type per LPAR
򐂰 Maximum number of adapters of one type per pSeries 690 or per pSeries 670
򐂰 Maximum combination of specific adapters
6
The boot capability from different disk technology depends on the operating system support. AIX supports these types
of disk for a boot device.
54
IBM ^ pSeries 670 and pSeries 690 System Handbook
The complete set of limitations are described in the PCI Adapter Placement
References, SA38-0538. This book is regularly updated and should be
considered as the reference for any questions related to PCI limitations.
Also see the IBM ^ pSeries 690 Configuring for Performance white
paper, found at:
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_config.html
We only mention two general rules of thumb here:
򐂰 All adapters that are supported in the RIO drawers are supported in the RIO-2
drawers, except three of them which use a 5V signaling:
– FC2751: S/390® ESCON® Channel PCI Adapter
– FC6206: PCI Single-Ended Ultra SCSI Adapter
– FC8396: IBM RS/6000 SP System Attachment Adapter
򐂰 In an RIO drawers, there is a maximum number of high speed adapters of 5
per planar, and 10 per drawer, while all the I/O slots of an RIO-2 drawers can
be populated with high speed adapters (for example, Gigabit Ethernet, Fiber
Channel, ATM or Ultra-320 SCSI adapters).
Two special configurations of I/O drawers are supported, which we mention even
though they are not recommended:
򐂰 A half-RIO drawer with only one I/O planar is available for an entry pSeries
670 with only one I/O drawer. In this case, only 8 SCSI disk slots and 10 I/O
slots are usable in the I/O drawer.
򐂰 A mixed I/O drawer where one of the I/O planars is an RIO planar, and the
other one is an RIO-2 planar. This configuration should only be used in very
special cases: when the customer needs one of the three I/O adapters which
are not supported with the RIO-2 planar, or when the customer upgrades a
half-RIO drawer into a full RIO drawer.
I/O drawer RAS
If there is an RIO failure in a port or cable, an I/O planar board can route data
through the other I/O connection and share the remaining RIO cable for I/O.
For power and cooling, each drawer has two redundant DC power supplies and
four high reliability fans. The power supplies and fans have redundancy built into
them, and the drawer can operate with a failed power supply or a failed fan. The
hard drives and the power supplies are hot swappable, and the PCI adapters are
hot plug.
All power, thermal, control, and communication systems are redundant in order
to eliminate outages due to single-component failures.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
55
2.4.2 I/O subsystem communication and monitoring
There are two main communication subsystems between the CEC and the I/O
drawers. The power and RAS infrastructure are responsible for gathering
environmental information and controlling power on I/O drawers. The RIO loops
are responsible for data transfer to and from I/O devices.
Power and RAS infrastructure
The power cables that connect each I/O drawer and the bulk power assembly
(BPA) provide both electricity power distribution and Reliability, Availability, and
Serviceability (RAS) infrastructure functions. The BPA is described in 2.5.1, “Bulk
power assembly” on page 73.
The primary I/O book has a BPC Y-cable connector that is used to connect
between two bulk power controllers (BPCs) on the bulk power assembly (BPA)
and the service processor using a Y-cable. Then two BPCs fan out the
connection to each of the I/O drawers, media drawers, fans, and Internal Battery
Features (IBFs). They compose the logical network (which we refer to as the
RAS infrastructure).
The RAS infrastructure includes:
򐂰 Powering all system components up or down, when requested. These
components include I/O drawers and the CEC.
򐂰 Powering down all the system enclosures on critical power faults.
򐂰 Verifying power configuration.
򐂰 Reporting power and environmental faults, as well as faults in the RAS
infrastructure network itself, on operator panels and through the service
processor.
򐂰 Assigning and writing location information into various VPD elements in the
system.
Note: It is the cabling between the RIO drawer and the BPA that defines the
numbering of the I/O drawer.
The power and RAS infrastructure monitors power, fans, and thermal conditions
in the system for problem conditions. These conditions are reported either
through an interrupt mechanism (for critical faults requiring immediate AIX
action) or through messages passed from the RAS infrastructure to the service
processor to RTAS. For more detailed information on the RAS features of both
the pSeries 670 and pSeries 690, see Chapter 5, “Reliability, availability, and
serviceability” on page 157.
56
IBM ^ pSeries 670 and pSeries 690 System Handbook
Remote I/O loop
The communication from the CEC to the I/O drawers is done over the remote I/O
link. This link uses a loop interconnect technology to provide redundant paths to
I/O drawers. RIO availability features include CRC checking on the RIO bus with
packet retry on bus timeouts. In addition, if a RIO link fails, the hardware is
designed to automatically initiate an RIO bus reassignment to route the data
through the alternate path to its intended destination.
The components involved in the creation of a loop are:
򐂰 RIO books, located in the CEC: The primary and secondary books have,
respectively, 2 or 4 pairs of RIO ports.
򐂰 RIO planars located in the I/O drawer: All drawers have 2 pairs of ports: one
on each planar.
򐂰 RIO cables: Cables are different to plug into an RIO port or an RIO-2 port.
There are two types of loops, RIO and RIO-2, depending on the technology used
in these components.
򐂰 RIO Loop connections operate at 500 MHz, and connect to the CEC using the
RIO books (FC6404 or 6410).
򐂰 RIO-2 Loop connections operate at 1 GHz, and connect to the CEC using the
RIO-2 books (FC6418 or 6419).
There are two modes of loop operation:
򐂰 Single-loop mode: The two I/O planars of an I/O drawer belongs to the same
loop, which is connected to one pair of ports of an I/O book
򐂰 Dual-loop mode: Each I/O planar of an I/O drawer belongs to one loop, which
connects to one pair of ports of an I/O book. The two loops connected to one
I/O drawer must connect to pairs of ports in the same IO book.
RIO books only support single-mode loops, while RIO-2 books support both
single-mode and dual mode loops.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
57
Figure 2-21 presents the cabling pattern that is supported for configuration of
loops between RIO books and RIO planers.
Cabling pattern #1
RIO book
GX-RIO
Single loop
GX-RIO
RIO book
2 x RIO planar
I/O drawer
RIO planar
RIO planar
GX-RIO
GX-RIO
RIO book/planar
RIO - RIO cable
Figure 2-21 RIO loops supported configurations
Figure 2-22 on page 59 presents the four cabling patterns that are supported for
configuration of loops between RIO-2 books and RIO or RIO-2 planers.
58
IBM ^ pSeries 670 and pSeries 690 System Handbook
Cabling pattern #2
RIO-2 book
GX-RIO-2
Single loop
RIO-2 book
GX-RIO-2
2 x RIO planar
I/O drawer
RIO planar
RIO planar
GX-RIO-2
GX-RIO-2
Cabling pattern #3
RIO-2 book
Dual loop
GX-RIO-2
RIO-2 book
2 x RIO-2 planar
GX-RIO-2
I/O drawer
RIO-2 planar
RIO-2 planar
GX-RIO-2
GX-RIO-2
Cabling pattern #4
RIO-2 book
GX-RIO-2
Single loop
RIO-2 book
GX-RIO-2
2 x RIO-2 planar
I/O drawer
RIO-2 planar
RIO-2 planar
GX-RIO-2
GX-RIO-2
Cabling pattern #5
RIO-2 book
GX-RIO-2
Dual loop
RIO-2 book
RIO and RIO-2 planar
GX-RIO-2
I/O drawer
RIO-2 planar
RIO planar
GX-RIO-2
GX-RIO-2
RIO book/planar
RIO-2 book/planar
RIO - RIO cable
RIO - RIO-2 cable
RIO-2 - RIO-2 cable
Figure 2-22 RIO-2 loops supported configurations
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
59
Table 2-8 lists the maximum number of RIO drawers and the maximum data rate
that can be reached for each of the cabling pattern presented in Figures 2-21 and
2-22.
Table 2-8 Maximum configuration for each cabling pattern
Cabling
Pattern
Max # of RIO
Drawers for
pSeries 690
Configuration
Max # of RIO
Drawers for
pSeries 670
Configuration
Maximum I/O
burst rate per
RIO Drawer
Maximum I/O
sustained
rate per RIO
Drawer
#1
8
3
2 GB/s
800 MB/s
#2
8
3
2 GB/s
800 MB/s
#3
7
21
8 GB/s
2.6 GB/s
#4
8
3
4 GB/s
1.6 GB/s
3 GB/s
2 GB/s
#5
8
1
2
1. Only two pairs out of four can be used on the Secondary I/O book of a pSeries 670.
An example of RIO cabling for the first three I/O drawers with IBF configuration is
shown in Figure 2-23 on page 61. For a complete set of RIO cabling diagrams,
refer to Appendix G, “Subsystem Positioning and Cabling” of IBM ^
pSeries 690 User’s Guide, SA38-0588 and IBM ^ pSeries 670 Installation
Guide, SA38-0613.
60
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 2-23 RIO connections for three I/O drawers with IBF configuration
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
61
The pairs of RIO book ports in the system are numbered from 0 to 13 as shown
in Figure 2-12 on page 41 (called GX-RIO in figure). In each book, the RIO port
pairs are named A0&A1, B0&B1, C0&C1, D0&D1, as shown in Figure 2-15 on
page 48. The relation between the pair number and the physical port pair is given
in Table 2-9.
Table 2-9 Physical I/O book ports and system logical port numbers
I/O Book
Pair Name
Pair Number
I/O Book
Pair Name
Pair Number
Primary - 0
A0&A1
0
Secondary - 2
A0&A1
10
B0&B1
1
B0&B1
11
N/A
C0&C1
4
N/A
D0&D1
5
A0&A1
6
Secondary - 1
A0&A1
2
Secondary - 3
B0&B1
3
B0&B1
7
C0&C1
8
C0&C1
12
D0&D1
9
D0&D1
13
Of course, if I/O books are not present, the corresponding pair numbers do not
appear in the system. For example, a system with only the Primary I/O book and
the Secondary I/O book 2 contains only the pairs numbered 0, 1, 4, 5, 10, 11.
The cabling rule for the I/O book port pairs is to connect the I/O loop pairs in the
following order: 0, 1, 4, 5, 10, 11, 12, 13, 6, 7, 2, 3, 8, and 9, without skipping any
installed pair.
There is no mandatory rule for the order to connect the I/O drawers. However,
pairs 0, 1, 4, 5, 10, 11, 12, 13 support both single-loop and dual-loop modes,
while pairs 6, 7, 2, 3, 8, and 9 support only dual-loop mode. Therefore, we
recommend you first connect to the CEC the drawers configured in single-loop
mode. For initial order, we also recommend that if the configuration contains
single-loops, you assign them to the first I/O drawers, to connect the first I/O
drawers to the first pairs of I/O book ports. Another consequence of the pairs 2,
3, 8, and 9 supporting only dual-loop mode is that the last GX bus position
(secondary book 1) is never used in a configuration containing only RIO drawers.
Table 2-10 presents for each combination of installed MCM and I/O books, the
lists of I/O ports that are usable on the CEC. Please note that this table is
cumulative top to bottom, and left to right: a system in row marked “MCM1” also
contains the MCM0 and MCM 2, and a system in column marked “Secondary-3”
also contains the primary book and the secondary-2 book. The entry for each
62
IBM ^ pSeries 670 and pSeries 690 System Handbook
combination contains two lines: the first line indicates the usable ports supporting
both single-loop and dual-loop modes, while the second line lists the port that
only supports dual-loop mode. Ports in the second lines cannot be used to
connect RIO drawers, but only RIO-2 drawers.
For example, with two MCMs and three I/O books, ports 0, 1, 4, and 5 are usable
to connect single-loops, and ports 6 and 7 are usable to connect dual-loops.
Table 2-10 Available I/O ports versus installed MCM and Books
MCM
position
Primary - 0
Secondary- 2
Secondary - 3
Secondary 1
0, 1
N/A
N/A
N/A
N/A
N/A
N/A
N/A
0, 1
0, 1, 4, 5
0, 1, 4, 5
N/A
N/A
N/A
6, 7
N/A
0, 1
0, 1, 4, 5, 10, 11
0, 1, 4, 5, 10, 11
0, 1, 4, 5, 10, 11
N/A
N/A
6, 7
6, 7, 2, 3, 8, 9
0, 1
0, 1, 4, 5, 10, 11
0, 1, 4, 5, 10, 11, 12, 13
0, 1, 4, 5, 10, 11, 12, 13
N/A
N/A
6, 7
6, 7, 2, 3, 8, 9
MCM 0
MCM 2
MCM 1
MCM 3
Figure 2-24 graphically presents the same information. It depicts four simplified
rear views of the CEC, when 1, 2, 3, or 4 MCMs are installed, and shows in each
case which pairs of ports can be used.
When installing additional MCM through MES to an existing system, the set of
available port pairs will be modified. For example, a 2-MCM and 3-I/O Book
systems can use pairs 1, 2, 4, 5, 6, and 7. After installation of a third MCM, it can
use pairs 0, 1, 4, 5, 10, 11, 6 and 7. If all pairs were used in the initial system,
there is no need to re-cable the I/O loops after installation of the third MCM, even
though the newly available pairs 10 an 11 have a higher cabling priority than
pairs 6 and 7.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
63
0
pair 1
0
1
1
1
1
0
0
1
1
1
1
1
0
0
1
1
Figure 2-24 Number of usable RIO or RIO-2 ports
IBM ^ pSeries 670 and pSeries 690 System Handbook
0
0
1
1
0
1
1
0
0
1
1
0
0
1
1
0
2
3
1
pair 1
1
0
0
pair 0
pair 3
1
pair 2
pair 9
pair 8
pair 2
useable RIO or RIO-2 ports
64
1
1
0
0
1
1
0
0
1
1
pair 12
1
1
pair 13
0
0
1
pair 6
0
0
0
pair 7
1
0
0
pair 5
1
1
0
pair 4
0
pair 11
0
1
pair 10
1
0
2
pair 12
1
pair 5
0
pair 1
1
0
0
pair 0
0
pair 13
1
pair 3
1
pair 4
pair 9
pair 8
pair 11
pair 10
0
0
4 MCM configuration
3 MCM configuration
0
1
pair 5
0
pair 4
0
pair 12
0
1
0
pair 1
1
0
pair 13
0
pair 0
2 MCM configuration
pair 0
1 MCM configuration
I/O cabling conclusion
An important change that is brought along with the new features announced for
pSeries 670 and pSeries 690 in May 2003 relates to the I/O loop cabling rules.
Before this date, the I/O loop had to be cabled according to the rules specified in
the IBM ^ pSeries 670 Installation Guide, SA38-0613 and the IBM
^ pSeries 690 Installation Guide, SA38-0587. Now, these rules are only
guidelines, and the I/O loops only need to be cabled in the next available I/O port
pair on the I/O books.
However, to simplify the management of the server we strongly recommend that
I/O loops be configured as described in the pSeries installation guides, and to
only follow a different order when absolutely necessary, for example, when
required by an MES installation or when using a mix of RIO and RIO-2 drawers.
In any case, it becomes extremely important for the management of the system
to keep an up-to-date cabling documentation of your systems, because it may be
different from the cabling diagrams of the pSeries installation guides.
Examples of RIO loop cabling and associated performance figures are given in
Appendix B, “I/O loop cabling and performance” on page 217.
2.4.3 I/O drawer physical placement order
The I/O drawer physical placement order in the pSeries 690 depends on whether
the system is equipped with the Internal Battery Feature (IBF).
The first three I/O drawers have a fixed placement order, regardless of the
system configuration in the base rack. They occupy the drawer positions 1, 2,
and 3, as shown in Figure 2-25.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
65
Without IBF
With IBF
Blank Space
Blank Space
IBF Position 3
9: I/O drawer #8
9: I/O drawer #8
IBF Position 1
IBF Position 2
8: I/O drawer #7
Blank Space
8: I/O drawer #7
4: I/O drawer #4
7: I/O drawer #4
1: I/O drawer #1
7: Blank Space
1: I/O drawer #1
6: I/O drawer #6
2: I/O drawer #2
6: I/O drawer #6
2: I/O drawer #2
5: I/O drawer #5
3: I/O drawer #3
5: I/O drawer #5
3: I/O drawer #3
Figure 2-25 I/O drawer and IBF placement positions (pSeries 690)
With IBF configuration, the fourth I/O drawer occupies drawer position 7 in the
expansion rack, as shown in the left side of Figure 2-25. Additional I/O drawers
are placed on the expansion rack in sequence from the drawer position 5 to 9,
but the drawer position 7 is skipped.
If the IBF is ordered, it is placed on the base rack between the media drawer and
the first I/O drawer (shown as IBF position 1 and IBF position 2; see Figure 2-25).
Up to four batteries can be placed on the base rack. You can also install two
additional batteries on the expansion rack between drawer positions 8 and 9
(shown as IBF position 3; see Figure 2-25).
Without IBF configuration, the fourth I/O drawer can be placed on the base rack.
It is placed between the media drawer and the first I/O drawer, as shown in the
right side of Figure 2-25. In this case, drawer position 7 in the expansion rack is
never used, as shown in the right side of Figure 2-25.
As already mentioned, the pSeries 670 only offers a maximum of three I/O
drawers and the IBF feature. Therefore only one frame is needed. Figure 2-26
shows the maximum I/O drawer configuration of pSeries 670.
66
IBM ^ pSeries 670 and pSeries 690 System Handbook
W ith or without IBF
IBF Position 1
IBF Position 2
1: I/O drawer #1
2: I/O drawer #2
3: I/O drawer #3
Figure 2-26 I/O drawer and IBF placement positions (pSeries 670)
The service processor is aware of the presence of the IBF, using the same
mechanism explained in “Power and RAS infrastructure” on page 56.
Note: Do not misunderstand the difference between a drawer position number
and an I/O drawer number. The drawer position number represents a
physically fixed position in the base or expansion racks. An I/O drawer number
represents the order of I/O drawers configured on the system.
All components of a pSeries 670 or a pSeries 690 can be identified by their
physical location code. The format of this code is:
U<rack_number>.<rack_position>[<optional location info>]
Where:
<rack_number> is 1 for the first rack, 2 for the second rack.
<rack_position> is the lowest position occupied by the device in the rack,
from 1 at the bottom of the rack to 42 at the top of the rack.
<optional location info> indicates a position within the device.
Figure 2-27 and Figure 2-28 present all physical location codes existing in
pSeries 670 or a pSeries 690 frames. For detailed information, refer to Chapter 1
of IBM ^ pSeries 670 Service Guide, SA38-0615 and IBM ^ pSeries
690 Service Guide, SA38-0589.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
67
Figure 2-27 Physical location in frame 1 of pSeries 670 and pSeries 690
68
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 2-28 Physical location in frame 2 of pSeries 690
Table 2-11 shows the relationship between I/O drawer positions and physical
location codes.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
69
Table 2-11 Physical location code of drawer position number
Drawer position number
Physical location code
pSeries 670
pSeries 690
1
U1.9
Yes
Yes
2
U1.5
Yes
Yes
3
U1.1
Yes
Yes
4
U1.13
No
Yes
5
U2.1
No
Yes
6
U2.5
No
Yes
7
U2.9
No
Yes
8
U2.13
No
Yes
9
U2.19
No
Yes
2.4.4 Media drawer
The media drawer provided in the base configuration contains an operator panel,
a 1.44 MB diskette drive and two sets of two removable media bays. The
operator panel and the diskette drive are each connected to the service
processor book in the CEC using point-to-point cables, and are powered from the
CEC standby power. Each set of media devices is powered and logically driven
by an I/O planar in the first I/O drawer, as shown in Figure 2-29.
70
IBM ^ pSeries 670 and pSeries 690 System Handbook
Front power connection
Media drawer (rear view)
Rear power connection
Front SCSI connection
Operator
panel
connection
Rear SCSI connection
First I/O drawer (rear view)
Diskette
drive
connection
Figure 2-29 Media drawer power and SCSI connection
The media drawer is split into two separate sections, as shown in Figure 2-30.
The front section houses the system operator panel, a diskette drive, and two
media bays. The rear section provides space for two additional media devices.
The front and rear sections of the media drawer are in two separate SCSI buses,
and must be connected to separate SCSI PCI adapters on the I/O drawers.
Therefore, these devices configured in the front and rear bays of the media
drawer can either be in the same partition or in different partitions.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
71
1U
Rear Media
Front View
Diskette
Op
Panel
Front Media
Front Media
Rear View
Op
Panel
Diskette
Operator panel and
diskette cables
to service processor
Rear Media
SCSI cable to SCSI adapter
and Power cable to I/O drawer
Figure 2-30 Media drawer
In order to use an IDE DVD-ROM drive (FC 2634) in the media drawer, the IDE
media to LVD SCSI interface bridge card (FC 4253) must be attached to each
IDE DVD-ROM drive. The IDE DVD-ROM drive is treated as a SCSI CD-ROM
drive on AIX.
Note: The IDE DVD-ROM drive (FC 2634) is supported on AIX 5L Version 5.2
and later.
2.5 Power subsystem
The power subsystem on the pSeries 670 and pSeries 690 are identical and
provide full redundancy. The power subsystem consists of redundant bulk power
assemblies, bulk power regulators, power controllers, power distribution
assemblies, DC power converters, and associated cabling.
72
IBM ^ pSeries 670 and pSeries 690 System Handbook
2.5.1 Bulk power assembly
The bulk power assembly (BPA) is the main power distribution unit for the
pSeries 670 and pSeries 690. The redundant bulk power assembly converts AC
input to DC and distributes power at 350 V to each drawer where conversion is
made to the required chip level. It is composed of several power converters and
regulators. Two BPAs are mounted in front and rear positions and occupy the top
of the rack as an eight EIA height drawer unit (see Figure 2-31). They are
powered from separate power sources via separate line cords.
Each BPA consist of a stack of components (from bottom to top):
򐂰 Three Bulk Power Regulator (BPR)
򐂰 One Bulk Power Controller (BPC)
򐂰 Three Bulk Power Distributor (BPD). The first BPD must be installed on all
systems. The second and third are optional, depending on the number of I/O
drawers installed in the system.
Figure 2-31 Power subsystem locations in BPA
The BPA distributes power to the I/O drawers and the DCA in the CEC via
appropriate power cables, plugged into the BPC and BPD connectors (The
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
73
number of the DCA in the CEC depends on the number of installed MCMs and
I/O books).
Constant power monitoring hardware assists in the detection of failures in the
power subsystem that require maintenance. The power supplies, power
converters, and cables can be serviced without system disruption. See
Figure 2-31.
The BPA also provides a major function in the RAS infrastructure, as explained in
“Power and RAS infrastructure” on page 56.
2.5.2 Internal battery feature
The internal battery feature (IBF) provides emergency power in case of a power
outage. The pSeries 670 and pSeries 690 incorporate an Early Power Off
Warning (EPOW) capability that assists in performing an orderly system
shutdown in the event of a sudden power loss. The IBF protects against power
line disturbances and provides sufficient power to enable the system to shut
down gracefully in the event that the power sources fail.
Up to two IBF enclosures can be installed in the base rack configuration. One
additional IBF enclosure can be installed in an expansion rack (see Figure 2-25
on page 66).
The hold-up time of the IBF will vary considerably with conditions present at the
time of the power loss. Ambient temperature, age, and past use of the battery
can affect the discharge time.
For more detailed information about the IBF hold-up times for the pSeries 670
and pSeries 690, see the 7040 sections in Chapter 1, “Physical Characteristics of
Systems” of the publication Site and Hardware Planning Information, SA38-0508.
Attention: The document Site and Hardware Planning Information, SA38-0508
is updated very often. The version of this document which includes information
related to the related to the pSeries 670 and pSeries 690 models announced
in May 2003 is version 18. If you have an old version of this document, please
download the latest version from the IBM Web page:
http://publib16.boulder.ibm.com/pseries/en_US/infocenter/base/HW_site_hardware.htm
74
IBM ^ pSeries 670 and pSeries 690 System Handbook
2.5.3 Cooling
Several fans and blowers provide air cooling for the pSeries 670 and pSeries
690. The power supplies, the CEC, and each I/O drawer have a group of
redundant cooling devices. They operate at normal speed when all the cooling
devices are operational, and increase their speed if one device stops working in
order to maintain the air flow constant and provide proper cooling.
The CEC is cooled by the following devices:
򐂰 Four high-pressure, high-flow blowers in the base of the CPU cage that pull
air down the front of the cage through the memory cards and exhaust air up
the rear of the cage through the power converters and RIO books.
򐂰 Two additional identical blowers in the front of the cage blow air directly down
on the MCMs and L3 modules, which is then exhausted out the top rear of the
cage.
Each fan or blower is referred to as an Air Moving Device, which consists of a
Motor/Scroll Assembly (MSA) and a Motor Drive Assembly (MDA).
򐂰 The MSA consists only of a motor, wheel, and metal housing.
򐂰 The MDA has two 350 V and communication inputs (one from each BPA) and
a three-phase output designed to drive a brush-less DC motor using back
EMF commutation control. The power and RAS infrastructure is used to
control and activate the Air Moving Devices through the communication ports
(see 2.4.2, “I/O subsystem communication and monitoring” on page 56).
A micro controller within the MDA monitors and controls the motor drive and
provides precise speed control capability. Each Air Moving Device can draw
15-300 watts of power, depending on speed setting and back pressure, with the
maximum power for a set of two blowers being 500 watts. The Air Moving Device
design is identical to that of the Air Moving Device used in the zSeries servers.
All Air Moving Devices from the CEC can be removed and replaced without
powering down the system.
Please note that each I/O drawer has the cooling devices that have the same
functionality within the CEC, and also offers N+1 redundancy.
2.6 IBM Hardware Management Console for pSeries
The IBM Hardware Management Console for pSeries (HMC) provides a standard
user interface for configuring and operating partitioning-capable pSeries servers.
The HMC provides a graphical user interface for configuring and operating single
or multiple partitioning-capable pSeries servers (also called managed systems).
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
75
Figure 2-32 shows an example of the graphical user interface on the HMC.
Figure 2-32 Graphical user interface on the HMC
The HMC consists of a 32-bit Intel-based desktop PC with a DVD-RAM drive
running the Linux operating system. The application environment, with a set of
hardware management applications for configuration and partitioning, is written
in Java. The applications are based on the object-oriented schema using the
Common Information Model (CIM), an industry standard sponsored by the
Distributed Management Task Force (DMTF). A CIM object manager acts as a
repository and database look-up for all managed objects.
The pSeries 670 and pSeries 690 provide two special asynchronous ports
dedicated for HMC attachment, HMC1 and HMC2, shown in Figure 2-15 on
page 48. The connectors on these ports are differentiated from standard
asynchronous ports by providing a 9-pin female connector versus the standard
9-pin male serial connector. The HMC and managed systems are connected by a
serial cable available in 6 meter and 15 meter lengths7.
The HMC allows a system administrator to do the following management tasks
on managed systems:
7
The 128-port asynchronous adapter (FC 2944) provides a distance solution, as explained in 3.2.11,
“HMC configuration rules” on page 93.
76
IBM ^ pSeries 670 and pSeries 690 System Handbook
򐂰 Creating and maintaining a multiple-partitioned environment
򐂰 Displaying a Virtual Terminal window for each operating system session
򐂰 Displaying operator panel values for each partition
򐂰 Detecting, reporting, and storing changes in the hardware conditions
򐂰 Controlling the powering on and off of the managed system
򐂰 Acting as the Service Focal Point for service representatives to determine an
appropriate service strategy and enable the Service Agent Call-Home
capability
򐂰 Activating additional processor resources on demand (Capacity Upgrade on
Demand)
The graphical user interface on the HMC is based on the AIX 5L Version 5.2
Web-based System Manager, which allows the management integration of other
HMCs or pSeries systems running AIX 5L Version 5.1 and 5.2. Except for IBM
customer support representatives, the native Linux interfaces are hidden from
the user and are not accessible. No Linux skills are required to operate the HMC.
For further information about how to use and manage HMC, refer to the following
publications:
򐂰 IBM Hardware Management Console for pSeries Installation and Operations
Guide, SA38-0590
򐂰 IBM Hardware Management Console for pSeries Maintenance Guide,
SA38-0603
Note: The HMC is a mandatory feature of the pSeries 670 and pSeries 690.
Chapter 2. Hardware architecture of the pSeries 670 and pSeries 690
77
78
IBM ^ pSeries 670 and pSeries 690 System Handbook
3
Chapter 3.
Using the IBM Configurator
for e-business
This chapter is intended to help IBM employees and IBM business partners to
prepare a pSeries 670 and pSeries 690 configuration for price quotation using
the IBM Configurator for e-business (also known as e-config).
򐂰 Section 3.1, “What’s new with e-config” summarizes the changes in the use
of the configurator due to the May 2003 announcements.
򐂰 Section 3.2, “Configuration rules for pSeries 670 and pSeries 690” details
information you should know before you start using the configurator.
򐂰 Section 3.3, “IBM Configurator for e-business (e-config)” presents the steps
performed in the configurator to create an initial order.
򐂰 Section 3.4, “Configuration examples” contains examples of initial orders,
feature conversions, and model conversions.
We assume that you are familiar with installation and basic use of e-config. More
information on the IBM Configurator for e-business is available at:
http://ftp.ibmlink.ibm.com/econfig/announce/index.htm
© Copyright IBM Corp. 2003. All rights reserved.
79
3.1 What’s new with e-config
To support the announcements of the new pSeries 670 and pSeries 690 features
in May 2003, a new Version 4.1.1 of e-config has been released.
Note: At the time of writing of this book, an internal test of Version 4.1.1 of the
IBM Configurator for e-business was used as shown in Figure 3-1. Changes or
enhancements to e-config may occur so that information in this chapter might
not be completely accurate over time. The version of e-config is the number
displayed in the Base field of the ECheck tool provided with e-config (see
Figure 3-1). It is not the version displayed if you click Help -> About
SCPortofolio in the configurator.
Figure 3-1 IBM Configurator for e-business version
There are a few changes in the way you will use this new version to create new
configurations or to prepare MES to an existing system, and especially:
򐂰 The I/O books on the CEC and the I/O planars on the I/O drawers are
available in two technologies (RIO and RIO-2). As a result, you must now
specify which kind of these devices you want, while in the previous version,
e-config would automatically configure the I/O books and planars for you.
򐂰 Because the I/O drawers can be connected using single-loop or dual-loop
mode, there is no longer a unique relationship between the number of
80
IBM ^ pSeries 670 and pSeries 690 System Handbook
installed I/O drawers and the number of required I/O books. You need to
specify your requirements and cannot only rely on default values provided by
the configurator.
Furthermore, e-config is not able to verify the consistency between requirements
and configured features, like the number of LPARs and number of disks or
Ethernet adapters. For example, you may want to use external disks in a 7133
SSA drawer as boot devices so a config with four internal disks and eight LPARs
is perfectly valid.
Be sure you understand the architectural constraints presented in Chapter 2,
“Hardware architecture of the pSeries 670 and pSeries 690” on page 17 and the
configuration rules presented in 3.2, “Configuration rules for pSeries 670 and
pSeries 690” before you start using e-config.
3.2 Configuration rules for pSeries 670 and pSeries 690
In order to properly configure a pSeries 670 or pSeries 690, some requirements
and constraints must be satisfied. For example, there are dependencies between
some hardware components: installation of a 1.7 GHz MCM implies that the L3
cache feature FC 4199 is also installed. This section explains the requirements
and options needed to configure the minimum configuration, as well as additional
components configuration.
Besides the correctness of the configuration, it is important to configure the
system that best meets the needs of the application. Requirements like
availability, performance, and flexibility may require additional components or
devices. For more information on how to best configure a pSeries 690, see the
IBM ^ pSeries 690 Availability Best Practices white paper and the IBM
^ pSeries 690 Configuring for Performance white paper, found at the
following URLs:
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_avail.html
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_config.html
The complete and up to date list of configuration guidelines, dependencies and
requirements can be found in the “Limitations” section of the pSeries 670 and
pSeries 690 sales manual, and in the guide PCI Adapter Placement References,
SA38-0538.
Some of the requirements do not leave a choice to the user of the configurator,
for example, all systems need one and only one clock (FC 5251). In this case,
these mandatory features are automatically added by the configurator, and we
will not detail the related configuration rules in the following sections.
Chapter 3. Using the IBM Configurator for e-business
81
A pSeries 670 or pSeries 690 server consists of several components which all
have a machine type. When using e-config, each of these components is
customized using a different wizard.
򐂰 One or two system rack: The 7040-61R wizard configures the options of the
rack itself (doors, …) but also of the Bulk Power Assembly and of the
Integrated Battery Feature.
򐂰 One Central Electronics Complex: The 7040-671 or 7040-681 wizard
configures the CEC options as well as the options in the Media Drawer.
򐂰 At least one and up to eight I/O drawer: The 7040-61D wizard configures the
adapters and internal disks, as well as the displays, mouse and keyboard that
can be attached to the LPARs.
򐂰 One or two HMC: The 7315-C01 or 7315-C02 wizard now configures the
HMC as a separate device.
Note: In the first versions of the pSeries 690, the HMC was configured as a
feature of the CEC (FC 7315 and FC7316). These features are no longer
available. On new systems, the HMC must be ordered as machine type
7315-C01 or 7315-C02. However, for compatibility issues with older
configuration, the CEC wizard still contains options for the HMC: keyboard,
mouse, power cables, … You should no longer use these options of the CEC
wizard. All options of the currently available HMC must be configured through
the HMC wizard.
The automatically configured features are not visible in the wizards. You will only
find them in the reports generated by e-config.
There are many dependencies between the different components of pSeries 670
and pSeries 690. In many cases, the number of LPARs and the set of adapters to
include in each LPAR will define the number of IO drawers to configure. The type
of these adapters (especially if they are high-speed adapters), and the
constraints on the supported number of concurrent adapters is likely to be the
first factor to decide on the number of I/O required drawers. This will also
influence the model of the I/O drawers to use and the type of I/O loops to
configure. The number of I/O drawers will decide on the number of racks to
install. The type of drawers and loops will decide on the type and number of I/O
books to install in the CEC. We therefore recommend that you start by
configuring all I/O drawers options, then the racks option, and finish by the CEC
wizard. The HMC can be configured independently from the other components.
This will minimize the number of times you need to revisit your configurations.
82
IBM ^ pSeries 670 and pSeries 690 System Handbook
3.2.1 Minimum configuration for the pSeries 670 and pSeries 690
The sales manual for the pSeries 670 and pSeries 690 describes a minimum
configuration including the following items:
򐂰 One 7040-671 Central Electronics Complex with 1 MCM and 4 GB memory
for the pSeries 670. The minimum processor configuration is one 4-way 1.1
GHz MCM.
򐂰 One 7040-681 Central Electronics Complex with 1 MCM and 8 GB memory
for the pSeries 690. The minimum processor configuration is one 8-way 1.1
GHz MCM.
򐂰 One 7040-61D I/O drawer with at least one network adapter per partition. The
minimum configuration for the pSeries 670 is an I/O drawer with only 1 planar
(10 PCI slots). The minimum configuration for the pSeries 690 is an I/O
drawer with both planars (20 PCI slots).
򐂰 One 7040-61R system rack with front and rear doors.
򐂰 A media drawer with one CD-ROM, DVD-RAM, or DVD-ROM drive.
򐂰 The bulk power assembly (BPA) and associated converters and cables.
The sales manual minimum configuration corresponds to the absolute lowest
price of an orderable server.
The IBM Configurator for e-business has a concept of default configuration. This
is the configuration you obtain when you create a server configuration without
selecting any options.
Note: In general, the Sales manual minimum configuration and the e-config
default configuration are different, even though they do not differ by many
features.
There are several possible reasons for the differences between the sales manual
minimum configuration and the e-config default configuration:
򐂰 The default configuration may recommend a feature configuration that gives
better performance than the cheaper alternative indicated in the sales
manual.
򐂰 The default configurations may contain feature options expected by most
customers that are not part of the cheapest configurations.
In the case of the pSeries 670 and pSeries 690:
򐂰 The default configuration is not a valid orderable configuration. It does not
contain the mandatory rack rear door, and you must specify if you need a slim
Chapter 3. Using the IBM Configurator for e-business
83
or an acoustic rear door. With the addition of this single option, the
configuration is valid.
򐂰 The default configuration contains language and power cable options
compatible with your e-config country (or region) setting, while the sales
manuals only contains generic options.
򐂰 The pSeries 670 minimum configuration is a server with one half RIO drawer
(only one RIO planar), while the default configuration contains one full RIO
drawer.
Appendix A, “Minimum and default configurations” on page 209 contains the
sales manual minimum configuration and the e-config default configuration of
pSeries 670 and pSeries 690.
The minimum pSeries 670 and pSeries 690 configurations require AIX 5L
Version 5.1 (5100-01 Recommended Maintenance Level with APAR IY39794) or
higher.
For configurations other than the minimal, the set of rules described in the
following section applies. For an overview of all available feature codes of the
pSeries 670 and pSeries 690, refer to the IBM Sales Manual.
3.2.2 LPAR considerations
Before defining the set of hardware devices that you want to be present in your
pSeries 670 or pSeries 690 system, you must pay attention to the constraints
implied by the use of logical partitions. If you want to use several LPARs, then
you must remember that:
򐂰 For up to 16 LPARs, you can use either RIO or RIO-2 Books in the CEC (FC
6404, 6410, 6418 or 6419). For instantiating from 17 to 32 LPARs, you must
only use RIO-2 Books (FC 6418 and 6419).
򐂰 Each LPAR must contain at least one processor, one bootable device, one
network adapter, and a minimum of memory:
– 1 GB per LPAR with a system firmware and HMC software release at a
level earlier than the 10/2002 level
– 256 MB per LPAR with a system firmware and HMC software release at
10/2002 level or later
򐂰 Each PCI adapter and each built-in SCSI adapter of the RIO drawers belongs
to at most one LPAR. There is no sharing of these resources between the
LPARs. As a consequence:
– If you use disks internal to the RIO drawers as boot devices, you need at
least one RIO drawer for each set of four LPARs.
84
IBM ^ pSeries 670 and pSeries 690 System Handbook
– The four disk of each four disk pack belongs to the same LPAR.
– All storage devices in the front of the media drawer belong to the same
LPAR. All storage devices in the back of the media drawer belong to the
same partition. At most, two partitions can use the internal CD-ROM,
DVD-ROM, DVD-RAM, and tapes devices.
Note: Even though you enter in e-config the number of LPAR you want to
instantiate on your server, the configurator does not check that the number of
configured hardware features will be sufficient to support the requested
LPARs. You have to it check by yourself.
A TCP/IP connection between the HMC and the LPARs is mandatory to use the
DLPAR feature, and recommended in all cases. We strongly recommend that you
provide this TCP/IP connectivity using a dedicated Ethernet network. Each LPAR
should contain dedicated Ethernet adapters for this connection.
3.2.3 Processor configuration rules
There are some rules that must be followed when configuring processors in a
pSeries 670 or pSeries 690 system, as part of the options of the 7040-671 or
7040-681 wizard. These include:
򐂰 All processors in a pSeries 670 or pSeries 690 must operate at the same
speed.
򐂰 Each installed MCM is supported with 128 MB of Level 3 cache. There is a
specific L3 cache FC for each MCM FC. e-config automatically inserts the
right L3 cache option for you.
򐂰 The mandatory programmable processor clock card (FC 5251) is
automatically inserted by e-config.
򐂰 If needed, the Processor Bus Pass Through Modules (FC 5257) are
automatically configured by e-config. pSeries 690 servers configured with two
or three MCMs must have the empty processor positions populated with
Processor Bus Pass Through Modules (FC 5257). No Processor Bus Pass
Through Module are required on pSeries 690 with one MCM or four MCMs, or
on pSeries 670.
򐂰 MCMs require one or several capacitor books (FC 6198) to operate. The
number of capacity books depends on the number of installed MCM, and
these features are automatically configured by e-config.
Chapter 3. Using the IBM Configurator for e-business
85
3.2.4 Memory configuration rules
There are some rules when configuring memory cards in the pSeries 670 and
pSeries 690 servers, using the CEC wizard. These include:
򐂰 The pSeries 690 has eight memory slots. Four memory slots utilize
inward-facing memory cards and four utilize outward-facing memory cards.
The inward-facing memory slots are utilized by the first and second MCM
positions, while the outward-facing memory slots support the third and fourth
MCM positions. The pSeries 670 only uses the four inward-facing memory
slots, because it only uses the first and second MCM position.
򐂰 The combinations of memory boards supported on pSeries 670 and pSeries
690 are detailed in 2.3.2, “Memory subsystem for pSeries 690” on page 29,
2.3.4, “Memory subsystem for pSeries 670” on page 42, and 4.2.4,
“Supported CUoD Memory configurations” on page 141.
򐂰 Memory boards are available in sizes: 4 GB, 8 GB, 16 GB, 32 GB, and 64 GB.
򐂰 The 4 GB and 8 GB at 575 MHz memory features are only available for the
1.7 GHz processor configurations.
򐂰 All the available memory slots should be populated with memory boards
which size are as closely balanced as possible. Only one memory increment
difference is allowed between two memory boards. Therefore, if you mix
different sized memory cards, the following combinations are allowed: 4 GB
and 8 GB, 8 GB and 16 GB, 16 GB and 32 GB, and 32 and 64 GB.
򐂰 Memory slots must be populated in pairs of equal size memory boards. The
only exception is for entry server pSeries 670 or pSeries 690 system with only
one MCM, which can be equipped with only one memory board in slot 0. The
pSeries 670 can have either one 4 GB board or one 8 GB board in position 0.
The pSeries 690 can have one 8 GB board in position 0.
򐂰 Servers using the 1.7 GHz processor should use 567 MHZ memory boards to
obtain best performance. Using 500 MHz memory on these servers results in
degraded memory subsystem performance.
򐂰 When the system is running in a partitioned environment, effective memory is
reduced due to partition page tables and translation control entry (TCE)
tables.
Table 3-1 on page 87 details the approximate reserved memory for page and
TCE tables, effective usable partition memory, and the maximum number of
partitions for various configurations. The configuration number in the header row
of columns 5 to 8 refer to:
86
Conf.A
Partitions with any version of AIX or Linux, with firmware and HMC
release levels earlier than the 10/2002 release level.
Conf.B
AIX 5L Version 5.1, Post-10/2002 firmware
IBM ^ pSeries 670 and pSeries 690 System Handbook
Conf.C
AIX 5L Version 5.2 or Linux, Post-10/2002 firmware
Conf.D
AIX 5L Version 5.1, Post-5/2003 firmware
Conf.E
AIX 5L Version 5.2 or Linux, Post-5/2003 firmware
Table 3-1 Physical memory size and number of allocatable partitions
Total
Memory
(in GB)
Approx.
Memory
Overhead
(in GB)
Approx.
Usable
Partition
Memory
(in GB)
Conf.A
Maximum
Partition
Number
Conf.B
Maximum
Partition
Number
Conf.C
Maximum
Partition
Number
Conf.D
Maximum
Partition
Number
Conf.E
Maximum
Partition
Number
(see 1, 2
and 3)
(see 1, 2
and 4)
(see 2,
and 5)
(see 1, 2,
6 and 7)
(see 1, 2,
and 7)
2G
.75 to 1
1 to 1.25
0 and 0
4 and 0
4
4 and 0
4
4GB
.75 to 1
3 to 3.25
2 and 0
12 and 0
12
12 and 0
12
8GB
.75 to 1
7 to 7.25
6 and 0
16 and 0
16
28 and 0
28
16GB
.75 to 1
15 to
15.25
14 and 0
16 and 0
16
32 and 0
32
24GB
1 to 1.25
22.75 to
23
16 and 0
16 and 0
16
32 and 0
32
32GB
1 to 1.5
30.5 to 31
16 and 0
16 and 0
16
32 and 0
32
48GB
1.5 to 2
46 to 46.5
16 and 1
16 and 1
16
32 and 1
32
64GB
1.5 to 2.25
61.75 to
62.5
16 and 2
16 and 2
16
32 and 2
32
96GB
2 to 3.5
92.75 to
94
16 and 4
16 and 4
16
32 and 4
32
128GB
2.5 to 4
124 to
125.5
16 and 6
16 and 6
16
32 and 6
32
192GB
3.5 to 5.75
186.25 to
188.5
16 and 10
16 and 10
16
32 and 10
32
256GB
4.5 to 7.5
248.5 to
251.5
16 and 14
16 and 14
16
32 and 14
32
320 GB
5.5 to 9.25
310.75 to
314.5
16
32 and 18
32
384 GB
6.5 to 11
373 to
377.5
16
32 and 22
32
Chapter 3. Using the IBM Configurator for e-business
87
Total
Memory
(in GB)
Approx.
Memory
Overhead
(in GB)
Approx.
Usable
Partition
Memory
(in GB)
Conf.A
Maximum
Partition
Number
Conf.B
Maximum
Partition
Number
Conf.C
Maximum
Partition
Number
Conf.D
Maximum
Partition
Number
Conf.E
Maximum
Partition
Number
(see 1, 2
and 3)
(see 1, 2
and 4)
(see 2,
and 5)
(see 1, 2,
6 and 7)
(see 1, 2,
and 7)
448 GB
7.5 to
12.75
435.25 to
440.5
16
32 and 26
32
512 GB
8.5 to 14.5
497.5 to
503.5
16
32 and 30
32
1. These columns contain two numbers. The first number corresponds to the maximum number of
partitions with memory less or equal to 16 GB and the second number corresponds to the maximum
number of partitions with memory greater than 16 GB.
2. All partition maximums are subject to availability of sufficient processor, memory, and I/O resources to
support that number of partitions. For example, a system with only eight processors can only support
a maximum of eight partitions.
3. These rules apply to systems running partitions with any version of AIX or Linux, if the firmware and
HMC release levels are earlier than the 10/2002 release level. The overall number of partitions must
be less than or equal to 16.
4. These rules apply to systems running partitions with AIX 5L Version 5.1 if the 10/2002 system
microcode update or later and HMC Release 3 or higher are used. The HMC partition profile option
for the Small Real Mode Address Region option should not be selected for AIX 5L Version 5.1
partitions. These numbers reflect the maximum when running only AIX 5L Version 5.1 partitions, but
AIX 5L Version 5.1 and AIX 5L Version 5.2 partitions can be mixed, and may allow for additional
partitions to be run (up to the maximum of 16).
5. These rules apply to systems running partitions with AIX 5L Version 5.2 (or later) or Linux, if the
firmware and HMC release levels are at the 10/2002 release level or later. The HMC partition profile
option for Small Real Mode Address Region should be selected for these partitions.
6. These rules apply to systems running partitions with AIX 5L Version 5.1 if the 5/2003 system
microcode update or later and HMC Release 3 or higher are used. The overall number of partitions
must be less than or equal to 32.
7. The I/O book FC6418 is required on pSeries 690 to support more than 16 partitions.
3.2.5 I/O books
You have to follow some rules when configuring the I/O drawers in a pSeries 670
or pSeries 690 with the 7040-681 or 7040-671 CEC wizard:
򐂰 All I/O books must use the same RIO or RIO-2 technology. Therefore, you can
configure either:
– One FC 6404, and zero to three optional FC6410, (one max for pSeries
670)
88
IBM ^ pSeries 670 and pSeries 690 System Handbook
– One FC 6418 and zero to three optional FC 6419 (one max for pSeries
670)
򐂰 By default, e-config will include one FC 6404 RIO book, that you can replace
with a FC 6418 RIO-2 book.
򐂰 If you configure POWER4+ MCM, you must manually configure FC 6418 and
FC 6419 RIO-2 book.
򐂰 If you need more I/O loops than what can be supported by the Primary I/O
book, you need to manually add the required number of secondary books.
򐂰 The rules explaining the constraints between the installed MCM and the I/O
books are described in detail in 2.3.3, “MCMs and GX slots relationship for
pSeries 690” on page 39 and 2.3.5, “MCMs and GX slots relationship for
pSeries 670” on page 44. The rules defining the relation between the number
of required I/O books and the number of required I/O loops are described in
2.4.2, “I/O subsystem communication and monitoring” on page 56.
3.2.6 Media drawer configuration rules
The media drawer is configured within the CEC wizard. It consists of two sections
and contains four media slots available for optional storage devices: The front
section contains two media bays, and the rear section contains the two other
media bays.
򐂰 One media drawer (FC 8692) is required for each pSeries 670 or pSeries 690
server and is automatically configured by default with one diskette drive and
one optional CD-ROM driver (FC-2624).
򐂰 One media device capable of reading CD-ROM media (either CD-ROM FC
2624, DVD-ROM FC 2634, or DVD-RAM FC 2623) is required for each
pSeries 670 or pSeries 690 system. You can replace the CD-ROM drive
configured by default with a DVD-ROM or DVD-RAM drive.
Note: In order to use IDE DVD-ROM drive (FC 2634) in the media drawer,
one SCSI to IDE Interface Bridge card (FC 4253) must be attached to each
IDE DVD-ROM drive.
򐂰 The 4 mm tape drive (FC 6158) is allowed in the rear bays of the media
drawer only if the operating environment is maintained at 24 degrees C (75.2
degrees F) or below and only on systems installed at 2134 meters (7,000 ft)
altitude or below.
򐂰 The DVD-RAM drive, CD-ROM drive, and DVD-ROM drive are limited to
systems without Primary Integrated Battery Backup (FC 6200) or Redundant
Integrated Battery Backup FC #6201) when installed in the Media Drawer (FC
8692) rear bay locations.
Chapter 3. Using the IBM Configurator for e-business
89
򐂰 The IBM 80/160 GB Internal Tape Drive with VXA Technology (FC 6120) is
limited to systems without Primary Integrated Battery Backup (FC 6200) or
Redundant Integrated Battery Backup (FC 6201) installed, Media Drawer (FC
8692) front bay locations only, and a maximum system ambient operating
temperature of 28 C (82.4 F) at a maximum operating altitude of 2134 m
(7000 ft). Lower altitude have higher maximum ambient operating
temperatures. Refer to the 7040 sections of Site and Hardware Planning
Information, SA38-0508, for additional details.
The power to the media drawer is provided by the first RIO drawer. The
connection of the media drawer storage options (CD, DVD Diskette) to the CEC
is provided by SCSI adapters in the first RIO drawer. The cables to provide these
connections are configured using the 7040-61D wizard:
򐂰 The connectivity between the front section of the media drawer and the first
media drawer is configured by default by e-config, which will add in the first
RIO drawer one power cable FC6179, one SCSI cable FC 2122 and one
Single Ended SCSI adapter FC 6206.
Note: The Single Ended SCSI adapter FC 6206 is a 5V adapter that
requires an RIO planar. If you want to only use RIO-2 drawers, you need to
change the default configuration to use a different SCSI adapter to attach
the front section of the media drawer.
򐂰 If you install more than two optional storage devices in the media drawer, the
rear section of the media will be used and therefore needs also to be
connected to the first I/O drawer. You have to specify a second SCSI adapter.
The second power cable FC6179 is automatically configured. If you add
another Single Ended SCSI adapter FC 6206, one SCSI cable FC 2122 is
automatically added to the configuration.
򐂰 You can also replace the default SCSI adapter by a PCI Dual Channel Ultra3
SCSI adapter (FC 6203).
򐂰 Each section of the media drawer needs to be connected to a different
adapter, so if you use the rear media slots, and you do not want to use the
Single Ended SCSI adapter, you must configure two PCI Dual Channel Ultra3
SCSI adapters. You cannot connect the two sides of the media drawer to the
two ports of one Dual Channel Ultra3 SCSI adapter. If you only configure one
SCSI adapter in I/O drawer 1, e-config will add a Single Ended SCSI adapter.
For each PCI Dual Channel Ultra3 SCSI adapter you specify, one SCSI cable
FC 2122 and one Converter cable VHDCI to P FC 2118 are automatically
added to the I/O drawer configuration.
90
IBM ^ pSeries 670 and pSeries 690 System Handbook
3.2.7 I/O drawer configuration rules
You have to follow some rules when configuring the I/O drawers in a pSeries 670
or pSeries 690 with the 7040-61D wizard, regarding the type of I/O planar to
chose depending on the I/O books installed in the CEC and the type of adapters
you need to install in the drawers. These rules have been described in detail in
2.4, “I/O subsystem” on page 51.
In addition, there are constraints that apply to the type, number and position in
the drawer of I/O adapters. These contraints are described in the publication PCI
Adapter Placement References, SA38-0538, and the pSeries 670 or pSeries 690
sales manual.
By default, e-config instantiates the first I/O drawer automatically for each new
pSeries 670 or pSeries 690 configuration. You then need to manually add any
extra drawer.
By default, each drawer contains four 4-pack Hot swap backplane (FC 6564) to
receive internal SCSI disks, the first drawer is populated with 2 disks of 36 GB
(FC 3158), while the other drawers do not contain any disks.
The internal SCSI adapters supporting the internal disk in the hot swap
backplane are configured by default. They are part of the I/O planars.
3.2.8 I/O loops and cabling
The RIO and RIO-2 cables are automatically added to the configuration of each
I/O drawer:
򐂰 Their type (RIO, RIO-2 or RIO-to-RIO-2) is defined by the types of I/O books
configured in the CEC, and the type of I/O planar configured in the I/O
drawers.
򐂰 Their number depends on the type of loop you choose: single-loop mode or
dual-loop mode.
The granularity of loop-mode is on a per drawer basis. For each I/O drawer you
configure, you find in the Storage tab of the wizard a checkbox called Prefer
Dual Loop. If you check this box, two pairs of RIO cables are configured for this
drawer.
The defaults settings are to use RIO books, drawers, and cables in single-loop
mode.
Chapter 3. Using the IBM Configurator for e-business
91
3.2.9 Graphics console configuration rules
The pSeries 670 and pSeries 690 can have dedicated graphics consoles for
each partition, including the Full System Partition. The configuration of the
optional graphics console is performed through the 7040-61D wizard.
Up to eight graphics consoles can be configured in a system, and each console
requires the following features:
򐂰
򐂰
򐂰
򐂰
One POWER GXT135P graphics accelerator (FC 2848)
One quiet touch keyboard: USB (FC 8800 or 8840)
One mouse: Business black with keyboard attachment cable (FC 8841)
One keyboard/mouse attachment card: PCI (FC 2737)
You can have a maximum of two of the GXT135P graphics accelerators and
keyboard/mouse attachment cards per PCI planar, and a maximum of eight per
system. Only one USB port is used to connect to the USB keyboard out of the
available four USB ports on the keyboard/mouse attachment card. The USB
mouse is daisy-chained from the USB keyboard.
Note: You cannot use this graphics console for the operating system
installation.
3.2.10 Rack and power units configuration rules
The racks, Bulk Power Assembly and optional Integrated Battery Features are
configured using the 7040-61R wizard.
When configuring the system rack, you must consider the following rules:
򐂰 All 7040-61R racks and expansion feature racks must have door assemblies
installed. The following door assemblies are available:
– A sculptured black front door with copper accent is automatically
configured for the primary Model 61R rack (FC 6070) and the expansion
rack (FC 6071), if installed.
– An acoustic rear door (FC 6075) is available for the primary rack or the
expansion feature rack. This door should be utilized for installations
requiring a quieter environment.
92
IBM ^ pSeries 670 and pSeries 690 System Handbook
Note: With the acoustic rear door, the hot air flowing out of the rack is
blown upward toward the ceiling, while with the slim line rear door, the
hot air is blown to the back of the rack. If your pSeries 670 or pSeries
690 servers are located in front of other machines requiring cooling, you
may need to order the rack with an acoustic rear door. Please refer to
the Site and Hardware Planning Information, SA38-0508, for
environmental characteristics.
– A slim line rack door (FC 6074) is available for the primary or expansion
rack feature. This door should be utilized for installations where system
footprint is the primary consideration.
Note: The racks are 2.02 meters high, and may not pass through doors
or fit in elevators during the delivery to a customer site. It is possible to
order the rack with its top shipped separately, so that it is only 1.65
meters high. This feature is not configurable though e-config. You must
create a valid rack configuration, and add an RPQ 8A1173 to the order.
The RPQ is only orderable with the initial factory order.
The power assembly subsystem and all its options (Bulk power regulators FC
6186, Power controller features FC 6187, Power Distribution Assembly FC6188)
are automatically configured with the right number of optional features to support
the hardware features configured in the CEC and I/O drawers. The only feature of
the power subsystem you need to specify is the line cord use to connect the
frame to your site power distribution system. A default power cord is configured
according to the country (or region) settings of the configurator.
The battery backup features (if ordered) supply power to the bulk power
regulators (FC 6186). The primary battery backup features (FC 6200) function
with the bulk power regulators located in the bulk power assembly in the front of
the rack. The redundant battery backup features (FC 6201) function with the bulk
power regulators in the bulk power assembly in the rear portion of the rack. If you
decide to order the battery option, you need to order as many batteries as there
are Bulk Power Regulators (FC 6186) configured in the racks. The cabling for the
battery is automatically configured.
3.2.11 HMC configuration rules
The HMC models which are now available, and all optional features of the HMC
must be configured using the 7315-C01 or 7315-C02 wizard. The HMC options
that are visible in the CEC wizard are only for compatibility with former models of
HMC which are no longer available.
Chapter 3. Using the IBM Configurator for e-business
93
The IBM Configurator for e-business includes by default with each HMC a
mouse, a keyboard, a graphical display, the attachment cable to one pSeries 670
and pSeries 690 host, and the required software. They do not need to be ordered
separately unless you want a different model than the default ones.
We strongly recommend that you order one optional Ethernet adapter. An
Ethernet connection between the HMC and each active partition on the
partitioning-capable pSeries server is required. This connection is utilized to
provide several system management tasks, such as dynamic logical partitioning
to each individual partition, and collection and passing of hardware service
events to the HMC from the partition for automatic notification of error conditions
to IBM.
For a serial connection between the HMC and the managed systems, the
following rules apply:
򐂰 The HMC provides two integrated serial ports. One serial port is required for
each pSeries server controlled by the HMC. An additional serial port is
required for modem attachment if the Service Agent call-home function is
implemented.
򐂰 If more than two serial ports are required on the HMC, an 8-port or 128-port
asynchronous adapter (FC 2943 or 2944) should be utilized.
򐂰 One HMC1 can manage up to twelve pSeries 670 or pSeries 690 systems,
with a maximum total of 64 logical partitions on these systems combined.
򐂰 If a redundant HMC function is desired, the pSeries 670 or pSeries 690 can
be attached to two separate HMCs.
򐂰 The 128-port adapter (FC 2944), in combination with a Remote
Asynchronous Node (FC 8137), can be used for a long distance solution
between HMC and a managed system. This will provide distances up to 1100
feet or 330 meters, while normal RS-232 connections allow up to 15 meters.
򐂰 Up to two 8-port and 128-port Asynchronous Adapters are allowed per HMC.
For use of an HMC in a cluster environment or with other pSeries models (630,
650, or 655), refer to the Sales Manual for exact limitation of the number of
supported server per HMC.
3.2.12 SP Cluster 1600 considerations
pSeries 670 or pSeries 690 can be configured as part of a Cluster 1600
(7018-160) using e-config, by selecting the RS Cluster 1600 option in the
1
The previous model of HMC (FC 7315 and 7316) only supports respectively up to four or eight
managed systems.
94
IBM ^ pSeries 670 and pSeries 690 System Handbook
“Product Line” section of the “Global Settings” panel when you start using the
configurator (see 3.3.1, “Initial order” on page 98 for details).
In addition to the wizards mentioned in the previous sections, you will have to use
the “Cluster Model 1600” wizard to specify the cluster options, which are the
number of nodes of each type you will include in your cluster. e-config uses this
number to check that your cluster complies with the limitations on the number of
Cluster 1600 supported nodes, that are fully documented in the publication IBM
^ Cluster 1600 Harware Planning, Installation and Service, GA22-7863.
Note: All limitations regarding maximum number of nodes are not checked by
e-config. If you configure a large or complex cluster, we recommend that you
manually check compliance against the information in IBM ^ Cluster
1600 Harware Planning, Installation and Service, GA22-7863.
A cluster must contain at least two nodes. If you plan to grow your cluster over
time, you may want to start with only one pSeries 670 or pSeries 690, and add
extra servers into the cluster later on. This is possible as long as you create at
least two LPARs in your server. The “Cluster Model 1600” wizard checks that
there are at least two servers in the cluster. Select either two features 7018-0008
7040 type server or 7018-0009 LPAR (Switched).
After you have configured the cluster, you can configure pSeries 670 or pSeries
690 system with the same wizards than for a standalone server.
Cluster 1600 has limitations that are documented in the publication RS/6000 SP
Planning Vol.1, Harware and Physical Environment, GA22-7863. For example,
you cannot attach a graphical display to an LPAR in a clustered environment.
Note: The e-config wizards for the pSeries 670 or pSeries 690 system
components are not aware that the servers are built as part of a cluster,
therefore, they do not check for any cluster related restrictions. You have to
check these restrictions manually.
If you were to configure a graphical display attached to an LPAR, e-config will not
flag an error or warning. It will consider the configuration as valid, even though it
is not a supported configuration.
Depending on whether you plan to manage your cluster with PSSP or CSM, you
will also need to configure a Control Workstation or a Management station.
These servers are not configured within the cluster. You need to add another
“Initial Order” to your configuration, specifying “pSeries and RS/6000” as the
product line, to chose the server you need.
Chapter 3. Using the IBM Configurator for e-business
95
The SP Switch (FC 8396) or SP Switch2 (FC 8397) adapters are “double-wide”
PCI adapters. When you configure then, e-config will automatically add the right
number of blind swap cassette for double-wide adapters (FC 4598).
The SP Switch (FC 8396) is a 5V PCI adapters that can only be installed in an
RIO drawer (with I/O planar FC 6563).
You must think of configuring Ethernet adapters in each LPAR for the SP
Management Ethernet network. If you use switched configurations, there are
limitations on the slots where these adapters can be installed: Slots 8, 9, 18 or 19
for an SP Switch2 configuration, and slots 1 though 7 or 11 through 17 for an SP
Switch configuration.
3.2.13 Upgrade considerations
If you have an installed pSeries 670 or pSeries 690 system, you may want to
increase its processing power. As of May 2003, you have three possibilities for
upgrading the system:
򐂰 MES: To install addition features, for example, to add an MCM or install RIO-2
books. This is the solution to use when your system limitations lies with the
I/O subsystem, while the CPU and memory are able to cope with your
applications workload, by replacing all RIO component with RIO-2
components.
򐂰 Feature conversion: To replace an installed component by a more powerful
one. This is available for MCM and memory replacements to go, for example,
from a 1.1 GHz system to a 1.5 GHz system.
򐂰 Model conversion, to convert a pSeries 670 system into a pSeries 690
system.
The complete list of all possibilities for any pSeries 670 and pSeries 690
component can be found in the sales manual. Here are a few rules related to
these upgrades:
򐂰 Feature conversions are implemented on a “quantity of one for quantity of
one” basis.
In a pSeries 670, you can replace:
– One 1.1 GHz 4-way MCM with one 1.1 GHz 8-way MCM
– One 1.1 GHz 4-way MCM with one 1.5 GHz 8-way MCM
– One 1.1 GHz 8-way MCM with one 1.5 GHz 8-way MCM
You cannot replace one 1.1 GHz 4-way MCM with one 1.5 GHz 4-way MCM.
In a pSeries 690, you can replace:
– One 1.3 GHz 4-way HPC MCM with one 1.3, 1.5 or 1.7 GHz 8-way MCM
96
IBM ^ pSeries 670 and pSeries 690 System Handbook
– One 1.1 GHz 8-way MCM with one 1.3, 1.5 or 1.7 GHz 8-way MCM
– One 1.3 GHz 8-way MCM with one 1.5 or 1.7 GHz 8-way MCM
– One 1.5 GHz 8-way MCM with one 1.7 GHz 8-way MCM
This is to say you can go from any pSeries 690 MCM configuration to a
configuration with the same number of any more powerful MCM type.
򐂰 Model conversion from pSeries 670 system into a pSeries 690 is only
available if the pSeries 670 has already two 8-way MCM installed. The
conversion allows transition from a 1.1 GHz pSeries 670 to a 1.1 GHz pSeries
690, or from a 1.5 GHz pSeries 670 to a 1.5 GHz pSeries 690.
During the conversion, some parts are replaced like the system planar (FC
9670), others are added like the Pass-through modules, and others are
carried over, like the MCM, caches, I/O drawers, …
򐂰 During a feature conversion replacing POWER 4 MCM with POWER4+ MCM,
the RIO books must be replaced with RIO-2 books.
For the features that are prerequisite to the processor upgrade from POWER
4 to POWER4+, e-config will either automatically replace them, or flag an
incompatibility and ask the user to change them.
3.3 IBM Configurator for e-business (e-config)
The IBM Configurator for e-business is an application that provides configuration
support for hardware, software, and peripherals associated with IBM ^
product lines that are available for marketing. Functions provided by e-config
include:
򐂰 The ability to create, save, restore, and print hardware, software, and
peripherals configurations.
򐂰 Hardware and software interaction for identifying prerequisite and
incompatibility conditions.
򐂰 Interactive support for re-entering product categories and continuous
modification and adjustments to the configuration.
򐂰 The ability to modify new or existing initial order, MES, or upgrade
configurations.
򐂰 The ability to modify an installed base prior to beginning MES or upgrade
configuration.
򐂰 Support for feature exchange and feature conversion.
򐂰 The ability to download and upload saved files to the Host/IBMLink™.
Chapter 3. Using the IBM Configurator for e-business
97
Note: Effective July 5th, 2002, the IBM Portable Configurator for RS/6000
(PCRS6000) is no longer supported. It has been replaced by e-config.
The e-config allows you to make valid configurations for pSeries 670 and pSeries
690 servers. The e-config checks for prerequisites and incompatibilities, and
often will correct them automatically. However, you will save time if, before using
e-config, you have an understanding of the constraints implied by the pSeries
670 and pSeries 690 server architecture, that are listed in:
򐂰 The IBM sales manuals for the latest information
򐂰 Section 3.2, “Configuration rules for pSeries 670 and pSeries 690” on
page 81
This section assumes that you have e-config already installed on your
workstation. It will be easier for you to understand this section if you can exercise
e-config on your display while you read this chapter.
In this section, we only address standalone pSeries 670and pSeries 690 servers.
We do not cover configuration of clustered servers.
For information on how to install and start e-config, please check the e-config
home page:
http://ftp.ibmlink.ibm.com/econfig/announce/index.htm
3.3.1 Initial order
Start e-config and select a blank portfolio. In the main menu, select Portfolio ->
Add Initial Order or click the Add Initial Order icon.
Figure 3-2 Add Initial Order icon
Enter the name and description of the configuration, select pSeries and
RS/6000 Systems, and click OK. A menu will appear with options to select the
geographic region and/or country, and customer requested arrival date (CRAD).
You can also specify pricing options on the second tab.
On the third tab, select pSeries and RS/6000 for Product Line. Select Initial for
Order Type and select either AIX pre-installed, AIX not required, or AIX but not
pre-installed in the AIX options box. Click Configure. A panel similar to
Figure 3-3 on page 99 appears.
98
IBM ^ pSeries 670 and pSeries 690 System Handbook
View tabs
Message Button
Portfolio
List
Catalog
List
Configuration
View
Wizards
List
Figure 3-3 The e-config main panel
On the Catalog list, search for pSeries 690 and click it to open the menu tree.
Double-click Server. You will see that the Wizards list has been updated with the
pSeries 690 basic components: The rack, the I/O drawer, and the server (CEC)
itself. Also, if you have selected the option to include AIX, the Software Global
Options and Software Products for the specified server will be present.
Figure 3-4 shows the physical placement of these components inside the rack. In
the Configuration view there is a graphical representation of the pSeries 690
components.
Chapter 3. Using the IBM Configurator for e-business
99
Figure 3-4 Graphical representation of pSeries 690 in configuration view
To edit the configuration and add or remove features, double-click the desired
component on the graphic in the Configuration view or in the Wizards list. Or
right-click the item in the Wizards list and select Edit, as shown in Figure 3-4. To
add more items of the same type, select Duplicate. A wizard is presented with
100
IBM ^ pSeries 670 and pSeries 690 System Handbook
the features available for that component. To duplicate the whole pSeries 670 or
pSeries 690 configuration, Duplicate Solution can be selected. Software
options, including AIX-enhanced software subscription options may be altered by
double-clicking Software Products: (7040-681: Rack Server 1). Follow the
dialog boxes from there and select the desired software components together
with the desired period of enhanced software subscription or software
maintenance.
CEC options
In the Wizards list, double-click 7040-681: Rack Server 1. The first panel of the
wizard (Products) enables you to change the name of the server; select the
operating system and specify the number of LPARs that the system will have.
Figure 3-5 pSeries 690 CEC wizard after double-clicking Rack Server 1
Chapter 3. Using the IBM Configurator for e-business
101
Note: The number of LPARs is for information purposes only. It does not
change features in the system, and does not prevent the customer from
altering the number specified originally.
You can configure the number and type of MCMs, I/O books and memory cards
in the System tab (see Figure 3-6).
Unless you order CuOD MCM, you should only enter a figure in one line of the
processor option list to select one type of MCM. For the 1.1 GHz and 1.3 GHz
systems, where the feature code is different for the first MCM and the others,
e-config will automatically pick the right feature codes to satisfy the number of
MCM you specify.
On the same tab, in the System Options list, you choose the RIO books to
connect I/O drawers. When configuring one or two RIO drawers in single-loop
mode connected to an RIO book, no extra RIO book is needed. For other
configurations, refer to 2.4.2, “I/O subsystem communication and monitoring” on
page 56 to define the type and number of I/O books needed on your system,
then select the required FC 6404, 6410, 6418, and 6419.
Then configure the memory books you need. In systems with one or two MCMs,
you should only configure inward-facing memory cards. In systems with more
than two MCMs, you can configure both types.
Then press the Next > button. If there are inconsistencies between the various
choices you made, you will receive a warning or error message.
102
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 3-6 System tab for the pSeries 690
After clicking Next, the Storage tab allows you to configure the media options for
the media drawer. You can select between the available media devices; up to four
devices can be configured into the media drawer. Remember the media drawer is
split into two separate sections. The front section houses the system operator
panel, a diskette drive, and two media bays. Utilization of the rear section is
optional to provide space for two additional media devices if desired.
The front and rear sections of the media drawer must be connected to separate
SCSI adapters in the first 7040-61D I/O drawer. If you want to use the rear
section of the media drawer (for example, to assigned media devices to two
different partitions), you must configure two SCSI adapters in the first 7040-61D
I/O drawer (see “I/O drawer options” on page 106).
Chapter 3. Using the IBM Configurator for e-business
103
The Adapter Cards tab presents the cabling options to the HMC for a simple one
to one connection. If you want to manage several pSeries from the same HMC,
you have to select 8-Port or 128-Port Asynchronous adapters as options of the
HMC.
On the Additional tab, you can specify the language group for your system. You
should not use the entries in the Line Cords options. These features are only left
in this panel for consistency with the first generations of pSeries 690, to specify
the appropriate power cords for the HMC (FC 7316 and 7315) and its associated
displays. The HMC, its display and their power cords are now configured in a
different wizard. Power cabling for the 7040-681 Central Electronics Complex is
provided from the bulk power assemblies in the 7040-61R rack and is not
configured in the CEC wizard.
On the Additional tab, you can also specify orders for On/Off CoD.
After you select the options from the tabs, click Configure to update your
configuration.
If there are inconsistencies between the various choices you made, you receive a
warning or error message, and after clicking the OK button you are presented
again with the wizard panel. If the configuration is valid, the wizard panel closes
and you return to the main panel.
You can see the physical location of the components in the CEC. On the diagram
in the Configuration view, right-click the area corresponding to the CEC (the area
named Rack Server 1), and select Detailed Diagram. A figure similar to
Figure 3-7 is presented.
104
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 3-7 Detailed diagram for the pSeries 690 CEC
Note that the unoccupied MCM slots are covered by processor bus pass-through
modules, when the machine is configured with two or three MCMs. The e-config
automatically places the modules when necessary.
Similarly, you can see the physical location of the components in the media
drawer using the same procedure. Figure 3-8 shows a detailed diagram for the
media drawer.
Chapter 3. Using the IBM Configurator for e-business
105
Figure 3-8 Detailed diagram for the media drawer
I/O drawer options
Double-click the I/O Drawer 1 in the Configuration view. The first panel of the
wizard enables you to change the name of the drawer. The placement is done
automatically. The Target Host drop-down menu lets you select the server to
connect the I/O drawer. Currently, you should have a single server in the
selection list. Then, click the Next > button to display the Storage options, as
shown in Figure 3-9.
The Storage tab allows you to select the internal disks for the I/O drawer. You
can select up to a total of 16 disks. The 4-pack hot swap backplanes come
configured by default. Disks will be automatically placed by e-config in an
optimum and valid configuration; this cannot be changed in the e-config. Actual
physical installation will allow for exact placing of the disks.
In the Hardware list of the Storage tab, you can choose between RIO, RIO-2 or
mixed drawers, by selecting the I/O planar feature codes:
򐂰 Two FC 6563 for an RIO drawer,
򐂰 Two FC 6571 for an RIO-2 drawer, or
򐂰 One FC 6563 and one FC 6571 for a mixed drawer
In the first I/O drawer, you can also select the number of cables needed to
connect to the Media drawer (FC 2122). If you use have configured three or four
storage devices in the media drawer, you need to manually configure a second
SCSI drawer, and the two FC 2122 are automatically configured, or you can
specify two FC 2122 cables in the storage tab, and the two SCSI adapters are
automatically configured. However, if you select only two storage devices in the
media drawer, and you want to install one in front, and one in the back section,
you must manually specify a second FC2122 in the storage tab, as well as a
106
IBM ^ pSeries 670 and pSeries 690 System Handbook
second SCSI adapter in the Adapter Cards tab and a second Power cable (FC
6179) in the Additional tab.
Left of the Hardware list, you can see the Prefer Dual Loop checkbox
(highlighted by a circle in Figure 3-9). If you click this box and you have
configured a RIO-2 or mixed drawer, e-config will automatically configure two
pairs of IO cables of the right type. If you have configured an RIO drawer, clicking
this checkbox has no effect, since RIO drawers are not supported in dual-loop
mode.
Figure 3-9 Selection of I/O drawer storage and RIO cabling options
The Adapter Cards tab provides menus to configure the different adapters on
the I/O drawer. Figure 3-10 on page 108 shows the tab with the menu options.
Adapters will be automatically placed by e-config in an optimum and valid
configuration; this cannot be changed in the e-config. Actual physical installation
will allow for exact placing of the adapters.
Chapter 3. Using the IBM Configurator for e-business
107
Figure 3-10 Adapter tab for the I/O drawer
Upon selection of one adapter category, the adapters that belong to that
category are shown and you can specify the number of adapters on the I/O
drawer. The configurator places the adapters automatically into the I/O drawer in
a balanced placement whenever possible. Figure 3-11 on page 109 shows the
options when selecting the LAN adapter category.
108
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 3-11 Adapter tab with LAN adapter category selected
On the Graphics tab, you can select graphics adapters and monitors for the
pSeries 670 and pSeries 690 systems. The maximum is one per logical partition
and eight per system. Be sure that with every graphics adapter (FC 2848) the
following devices are also configured:
򐂰 One quiet touch keyboard: USB (FC 8800 to 8840), and one mouse: Business
black with keyboard attachment cable (FC 8841) from the Input Devices tab
And
򐂰 One keyboard/mouse attachment card: PCI (FC 2737, found on the Adapter
Cards tab, Async/USB)
On the Additional tab, you can specify in the first list the language group for the
publications provided with the hardware.
Chapter 3. Using the IBM Configurator for e-business
109
The Line Cords list of the Additional tab presents the optional power cords
needed if you configure a graphical display in the Graphics tab. When you add a
display, a power cord is automatically configured according to the e-config
country (or region) settings. You can then override this choice in the Line Cords
list.
In the Additional tab, you can also specify the number of power cables for the
media drawer. If you add more than two devices on the media drawer, then you
have to add another power cable. If you do not add the cables, e-config will
automatically add them for you when necessary. The power cable to the I/O
drawer (FC 6172) is automatically configured and should not be modified.
The Miscellaneous Equipment list of the Additional tab allows you to specify
Double Wide Adapters cassette for holding large PCI adapters. The number of
cassettes is automatically configured depending on the number of adapters you
selected in the Adapter Cards tab. You only need to change this number if you
want extra cassettes, for example to hold adapters you already own.
After you select the options from the tabs, click Configure to update your
configuration. The notebook will close and return to the main panel.
Similar to the CEC, you can see the physical location of the components in the
I/O drawer by right-clicking the area corresponding to the I/O drawer (the area
named pSeries 690 (7040-61D) Additional I/O Drawer: I/O Drawer 1) and
selecting Detailed Diagram. Figure 3-12 shows the diagram view. PCI adapters
and disks are placed in optimal configuration automatically by e-config. This
cannot be changed in the configurator. Exact placement of disks and adapters
can be done at physical installation time.
Figure 3-12 Detailed diagram of the I/O drawer
110
IBM ^ pSeries 670 and pSeries 690 System Handbook
You can now configure additional I/O drawers. If you have a need for several I/O
drawers with the same or similar configuration, the easiest solution is to
right-click on an already configured I/O drawer in the Wizards list (see
Figure 3-13), and then edit the new I/O drawer to modify its features.
Figure 3-13 Duplicating an existing drawer
If you want a totally different drawer configuration, you should instead
double-click on the I/O Drawer - (7040-6D1) entry in the Catalogs list, under the
pSeries 690 main entry:
Figure 3-14 Selecting a new I/O drawer
Chapter 3. Using the IBM Configurator for e-business
111
Rack options
Double-click the 7040-61R graphic in the Configuration view or right-click
7040-61R: Rack 1 in the Wizards list and select Edit. The first panel of the
wizard enables you to specify front and rear doors for the rack, and add an
additional 42U expansion rack. The acoustic rear door (FC 6075) should be
utilized for installations requiring a quieter environment.
When more than four I/O drawers are ordered, or four I/O drawers with integrated
battery backup (FC 6200), an expansion rack is automatically selected by
e-config. There is only one wizard that contains the features for both racks. The
expansion rack is indicated by the number “1” in the line (8691) - 42U Expansion
in the Options tab. The doors for both racks can be configured by double-clicking
any of the two racks. You have to specify two rear doors, either acoustic (FC
6075) or slim line (FC 6074).
The Additional tab lets you select the power cord options for the rack. A default
selection based on the country (or region) includes two cables. You can change
these cables and replace with them others. You must always specify two cables.
Optionally, you can select the optional battery features. Select the integrated
battery backup, primary (FC 6200). As an option for the primary battery backup,
you can select the redundant features (FC 6201). You can add up to three of
each battery. For each redundant battery, you must specify a primary battery.
Note: When primary integrated battery backup feature(s) are ordered, one is
required for each bulk power regulator (FC 6186) in the front bulk power
assembly. The e-config will notify you when more battery backup features are
necessary and select them automatically.
Hardware Management Console
The pSeries 670 and pSeries 690 servers need to be connected to an HMC. If
you already have an HMC installed that can be used to manage the server being
configured, you can skip this step. Otherwise, you need to configure an HMC. To
do this, click on the button on the left of the Hardware Management Console
entry in the Catalogs list. Then double click on the 7315-C01 or 7315-C02
Hardware Management Console entry. You are then presented with the HMC
wizard window, as presented in Figure 3-15.
112
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 3-15 HMC wizard
In the Products tab, you can change the HMC name.
Then skip to the Adapter Cards where you are presented by default with the
possibility of ordering extra Ethernet adapters. Click on the Adapter Category
pull-down menu to select Async/USB, as shown on Figure 3-16. You can then
select the attachment cables and optional Asynchronous adapters to connect the
HMC to the pSeries 670 and pSeries 690 servers.
The Graphics and Input Devices tabs allow you to change the display and
keyboard configured by default.
The Additional tab contains the default language and power cords
corresponding to the country (or region) settings, that you can override by
selecting other features.
Chapter 3. Using the IBM Configurator for e-business
113
Figure 3-16 HMC adapter options
You then click on the Configure button to add the HMC to the configuration.
Validating the configuration
Before validating the configuration, you should look for any information message,
warning or error reported by the configurator. Click the Message tab on the
Views tabs area, or click the Message button (see Figure 3-3 on page 99). A
panel similar to Figure 3-17 appears.
114
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 3-17 The error, warning, and information output of the configuration
If you right-click a message, a context-sensitive menu appears and gives you the
ability to execute actions on that message by selecting Jump to Wizard.
Figure 3-18 shows the possible actions for an error message.
Figure 3-18 Right-click the error message text and it opens a menu
If you select the option Detailed Message Text, a panel opens with more
information about the message, as shown in Figure 3-19. Click OK to return to
the messages panel.
Chapter 3. Using the IBM Configurator for e-business
115
Figure 3-19 Detailed error message
Another option is to select the Jump to Wizard action. If you select this, the
wizard panels related to that message will open, and you can perform the
required changes.
After correcting the configuration errors, right-click the message:
A final validation is required to complete this configuration
Click the Perform Final Validation icon (Figure 3-20).
Figure 3-20 Perform Final Validation icon
If no other errors are encountered, the system is validated and you can save the
output files and place the order.
On the Proposed tab, you can find a text description of the configuration, with all
the feature codes and prices for the hardware and/or software components.
To generate a price proposal and place an order, you have to save the files.
Select File -> Save or File -> Save As, or click the diskette icon in the upper left
corner of the panel.
3.3.2 Performing an upgrade
The configuration steps to perform an upgrade on a pSeries 670 or pSeries 690
system are the same steps used to configure a new one. The only thing that
changes is that instead of specifying an initial order you should specify
Upgrade/MES.
116
IBM ^ pSeries 670 and pSeries 690 System Handbook
From the main menu, select Portfolio -> Add Upgrade/MES or Restore
CFReport, or click the Add Upgrade/MES or Restore CFR icon.
Figure 3-21 Add Upgrade/MES or Restore CFR icon
If importing from a previously saved CFReport file, select Import local file.
Select the CFReport file to be imported, and after a moment the base
configuration is presented. You can then change components or add new
components to the configuration and proceed with the upgrade.
3.4 Configuration examples
In this section we provide the steps needed to make a complete configuration of
a pSeries 670 and a pSeries 690 system. We present several scenarios, going
from an initial order of a simple basic configuration to the update of a large
configuration with multiple I/O drawers and an expansion rack:
򐂰 Section 3.4.1 presents the initial order of a pSeries 670.
򐂰 Section 3.4.2 describes the initial order of a 2-frame pSeries 690.
򐂰 Section 3.4.3 explains how to perform a model conversion from a pSeries 670
into a pSeries 690 system.
򐂰 Section 3.4.4 presents feature conversion from a 1.3 GHz pSeries 690 to a
1.7 GHz pSeries 690.
In these sections we assume that the user is familiar with e-config and how to
create a new configuration. If any difficulties arise, we recommend that you see
3.3, “IBM Configurator for e-business (e-config)” on page 97.
3.4.1 Configuration example 1: pSeries 670 (16-way 1.1 GHz)
In this section we will prepare an order for a machine with the following
configuration:
򐂰
򐂰
򐂰
򐂰
򐂰
򐂰
򐂰
16-way 1.1 GHz processors
64 GB of memory
Eight 36 GB internal disks
Four logical partitions
Four Gigabit Ethernet adapters and four Fibre Channel adapters
One CD-ROM and one 4 mm DAT tape drive
One I/O drawer (RIO loops)
Chapter 3. Using the IBM Configurator for e-business
117
򐂰 One HMC with a 6 meter cable and 21 inch monitor, US keyboard, and mouse
򐂰 US Language and US power cords
򐂰 AIX 5L Version 5.1 with three years enhanced software subscription
To prepare the configuration reports, follow these steps:
1. In the Catalog list area, double-click pSeries 670 and double-click Server. You
will find a graphical representation of the pSeries 690 in your Configuration
view.
2. Double-click Rack Server 1. Select the AIX version and fill in the number of
required LPARs. Click Next. Go to the System tab and change the quantity of
the 4-way POWER4 processor from one to zero, and the quantity of the 8-way
POWER4 processor (FC 5256) to two. On the same tab, change the quantity
4 GB memory card, inward facing (FC 4196), from one to zero, then add four
16 GB memory cards, inward facing (FC 4183).
3. Go to the Storage tab and add a 20/40 GB 4 mm tape drive (FC 6158).
4. Click the Next button to move to the Adapter Cards tab. Change the quantity
of the console/host serial attachment cable, 15 M (FC 8121), from one to
zero, and change the quantity of the console/host serial attachment cable,
6 M (FC 8120), from zero to one. Click the Configure button to return to the
main panel.
5. Double-click I/O Drawer 1 in the Configuration view to open its wizard. Click
the Storage tab. Change the quantity of the 36.4 GB Ultra3 SCSI 10 K RPM
disk drive/1 inch - (FC 3158) from two to eight.
6. Click the Next button to move to the Adapter Cards tab and from the
Adapter Category menu; select LAN. Add four of the Gigabit Ethernet–SX
PCI-X adapters (FC 5700). Select the HiPerf category on the menu and add
four of the PCI Gigabit Fibre Channel adapters for 64-bit PCI bus (FC 6228).
Note: The FC 6228 card uses an LC connector. Most of the Fibre Channel
devices still use the SC connector. Therefore, we recommend that you add
the Fibre Channel Converter Cable, LC (M), to SC (F), 2M (FC 2456) for
each Fibre Channel adapter being configured.
7. Click the Configure button to return to the main panel.
8. Double-click 7040-61R in the Configuration view to open its wizard. Add one
rear door, acoustic, primary or secondary rack (FC 6075).
9. Go to the Additional tab and select a language group: US English (FC 9300).
Click the Configure button to return to the main panel.
118
IBM ^ pSeries 670 and pSeries 690 System Handbook
10.In the Catalog list area, double-click Hardware Management Console, then
double-click 7015-C02 Hardware Management Console. The HMC wizard
appears on the screen.
11.Change to the Graphics tab and add replace the IBM P76/P77 display with
one IBM P260/P275 color monitor, business black, and cable (FC 3628).
12.Go to the Input Devices tab and add one quiet touch keyboard: Business
black US English, #103P (FC 8800).
13.Click the Configure button to return to the main panel.
14.Click the Messages tab on the main panel and right-click the message. A
final validation is required to complete the configuration, so click the Perform
Final Validation icon (Figure 3-22).
Figure 3-22 Perform Final Validation icon
15.If you follow all these steps, your configuration will be valid, and you can save
the output file for processing. If you encounter any problems, check the steps
again and refer to the documentation.
3.4.2 Configuration example 2: pSeries 690 (24-way 1.3 GHz)
In this section we will configure a machine with the following resources:
򐂰 24-way1.3 GHz processors
򐂰 96 GB of memory
򐂰 Four I/O drawers, with RIO loops
򐂰 Forty 36 GB internal disks (ten in each I/O drawer)
򐂰 16 Gigabit Ethernet adapters 16 Fibre Channel adapters (four in each I/O
drawer) and two PCI ESCON Channel adapters (in the first I/O drawer)
򐂰 One DVD-RAM drive and one 4 mm DAT tape drive
򐂰 Two HMCs with 15 meter cable and 21 inch monitor, US keyboard, and
mouse
򐂰 Two Integrated Battery Feature options, one being principal and the other
redundant
򐂰 US Language and US power cords
򐂰 AIX 5L Version 5.1 and PCI ESCON Control Unit Connectivity software
Version 2.1
Chapter 3. Using the IBM Configurator for e-business
119
Follow these steps to perform the configuration:
1. In the Catalog list area, double-click pSeries 690 and double-click Server. A
graphical representation of the pSeries 690 appears in your Configuration
view.
2. Double-click Rack Server 1. Go to the System tab and change the quantity
of the 8-way POWER4 processor (FC 5242, 5252) from one to zero. Add
three of the 8-way POWER4 Turbo processors (FC 5244, 5254). On the same
tab, change the 8 GB card from one to zero, then add four 16 GB memory
cards, inward facing (FC 4183), and two 16 GB memory cards, outward facing
(FC 4184).
There are several ways to configure memory on the pSeries 670 and pSeries
690. The suggested configuration is recommended for best performance. For
more information on how to configure memory on the pSeries 670 and
pSeries 690, see Table 2-4 on page 36 and Table 2-6 on page 44,
respectively, or consult with the IBM ^ pSeries 690 Configuring for
Performance white paper, found at:
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_config
.html
3. Go to the Storage tab and change the quantity of the CD-ROM drive–32X
(Max) SCSI-2 (FC 2624) from one to zero. Add a 20/40 GB 4 mm tape drive
(FC 6158) and one DVD-RAM SCSI re-writable 4.7 GB black bezel (FC
2623).
4. Click the Next button to move to the Adapter Cards tab. Add another
console/host serial attachment cable, 15M (FC 8121).
5. Click the Configure button to return to the main panel.
6. Double-click I/O Drawer 1 to open its wizard.
7. Click Next to open the Storage tab. Change the quantity of the 36.4 GB
Ultra3 SCSI 10 K RPM disk drives/1 inch - (FC 3158) from two to 10.
8. Click the Next button to move to the Adapter Cards tab, and then from the
Adapter Category menu, select LAN. Add four of the Gigabit Ethernet–SX
PCI-X adapters (FC 5700). Select the HiPerf category on the menu and add
four of the PCI Gigabit Fibre Channel adapters for 64-bit PCI bus (FC 6228).
Add four Fibre Channel converter cables, LC (M) to SC (F), 2M (FC 2456),
and two S/390 ESCON Channel PCI adapters (FC 2751). Click the
Configure button to return to the main panel.
9. Right-click the I/O Drawer (7040-61D): I/O Drawer 1 in the components list
and select Duplicate. Since the configuration of I/O drawers 2, 3, and 4 is
identical, click the Configure button to return to the main panel. Repeat this
action to add the third and fourth I/O drawer.
120
IBM ^ pSeries 670 and pSeries 690 System Handbook
Note: Even if the I/O drawer configuration is different in terms of number
and type of adapters, you may want to use this Duplicate function. You just
need to change the adapters to the desired configuration after duplication.
Make sure the maximum number of adapters per system is not surpassed,
otherwise you will get an error message when you click Configure.
10.In order to accommodate the four I/O drawers and the batteries, we need an
additional rack. Double-click 7040-61R to open its wizard. Add two rear doors,
acoustic, primary or secondary rack (FC 6075). Add one front door (black) for
24 in 2M (42U) racks (FC 6071) and one 42U expansion (FC 8691).
11.Go to the Additional tab. On the Power Options list, select one of the
integrated battery backups, primary (FC 6200) and one of the integrated
battery backup, redundant (FC 6201) features. Click the Configure button to
return to the main panel. The e-config finds out that more battery backup
features are necessary and automatically selects them (one is required for
each bulk power regulator in the front bulk power assembly). Note that the
e-config will automatically configure the expansion rack when necessary.
12.Click the Messages tab on the main panel. An error message appears
because of a needed feature:
Rack Server 1: There must be exactly 1 Remote I/O Loop Adapter (6410) for
the current configuration
13.Right-click the message and select Jump to Wizard.
14.Go to the System tab and add one quad RIO adapter (FC 6410). Click the
Configure button to return to the main panel.
15.You should now add the required software for the ESCON adapter.
Double-click Software Products (pSeries 690 (7040-681): Rack Server 1)
and select the desired software. Configure the prerequisite software, such as
the PCI ESCON Channel Control Unit. You can use this panel to add
additional software. Select the software you want and click Next several
times, and finally Configure.
16.In the Catalog list area, double-click Hardware Management Console, then
double-click 7015-C02 Hardware Management Console. The HMC wizard
appears on the screen.
17.Change to the Graphics tab and add or replace the IBM P76/P77 display with
one IBM P260/P275 color monitor, business black, and cable (FC 3628).
18.Go to the Input Devices tab and add one quiet touch keyboard: Business
black US English, #103P (FC 8800).
19.Click the Configure button to return to the main panel.
Chapter 3. Using the IBM Configurator for e-business
121
20.Right-click in the Wizards list on the HMC entry, and select Duplicate to add a
second identical HMC.
21.Click the Messages tab on the main panel and right-click the message:
A final validation is required to complete the configuration
Then select Perform Final Validation.
22.The final graphical representation for this configuration should be similar to
Figure 3-23 on page 123.
122
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 3-23 A graphical representation of the final configuration
Chapter 3. Using the IBM Configurator for e-business
123
3.4.3 Model conversion from pSeries 670 to pSeries 690
The starting point for this example is the configuration prepared in 3.4.1,
“Configuration example 1: pSeries 670 (16-way 1.1 GHz)” .
Rather than using the method described in 3.3.2, “Performing an upgrade” , we
assume that the configuration is already loaded in e-config.
The steps involved in the model conversion are:
1. Start upgrading the configuration already loaded in e-config by right-clicking
the system in the Portfolios list, and select Start Upgrade. A new system is
created in the Portfolios list, and automatically displayed in the Configuration
view.
Figure 3-24 Starting to upgrade an existing configuration
2. Right-click the CEC, called Rack Server 1 in the diagram, and select
Upgrade.
124
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 3-25 Start upgrading the CEC.
The CEC wizard windows pops up, and the Products tabs contains a
Product Type list, where the current model, 7040-671, is highlighted. Select
the 7040-681 type (see Figure 3-26), and click the Next > button to change
the model.
Chapter 3. Using the IBM Configurator for e-business
125
Figure 3-26 Changing model to pSeries 690
3. Click on Configure to return to the Configuration view.
4. Switch to the Message view in the Configuration window. It contains errors
that need to be fixed.
Figure 3-27 Errors in the Messages tab
5. Double-click the error:
Each Server 1: Invalid number of AIX licenses configured.
126
IBM ^ pSeries 670 and pSeries 690 System Handbook
The detailed message indicates that the number of licenses originally ordered
for a 16-way server is no longer valid with a 32-way server. In the Wizards list,
right-click the Software Products wizard, and select Edit.
6. In the Software wizard, click on the Next > button, until you are presented with
the panel specifying the number of AIX licenses (see Figure 3-28). Add 16
licenses, and click Configure. The related error disappears from the
Message view.
Figure 3-28 Changing the number of AIX licenses
7. You can now perform the final validation and generate the configuration
reports.
3.4.4 Feature conversion from POWER4 to POWER4+
The starting point for this example is the configuration prepared in 3.4.2,
“Configuration example 2: pSeries 690 (24-way 1.3 GHz)” .
The desired configuration is a pSeries 690 with:
򐂰 Four 1.7 GHz MCM
Chapter 3. Using the IBM Configurator for e-business
127
򐂰 128 GB memory
򐂰 One RIO drawers and 4 RIO-2 drawers, with contents identical to the content
of the drawers in the base pSeries 690 configuration. The RIO drawer is
configured in single-loop mode, while the RIO-2 drawers are configured in
dual-loop mode.
This example shows that you can include in one configuration report feature
conversions (to replace existing processors for example) and MES to add new
features.
We assume that the base configuration is already loaded in e-config.
The steps involved in the model conversion are:
1. Start upgrading the base configuration right-clicking the system in the
Portfolios list, and select Start Upgrade (see the first step in 3.4.3, “Model
conversion from pSeries 670 to pSeries 690” for details). A new system is
created in the Portfolios list, and automatically displayed in the Configuration
view.
2. Right-click the CEC, called Rack Server 1 in the diagram, and select MES.
The CEC wizard windows pops up, opened on the Products tabs.
3. Switch to the System tab: In the Processor list, change the number of
features FC 5245 and FC 5254 to zero, scroll down, and change the number
of 1.7 GHz processor (FC 5246) to 4.
4. In the System Options list, reset the number of RIO books (FC 6404 and FC
6410) to 0, and select RIO-2 books (FC 6418 and FC 6419): two secondary
books are sufficient to connect the 9 I/O loops.
5. In the memory list, reset the number of 16 GHz memory card to zero (FC
4183 and FC4184), and select eight 567 MHz memory cards (FC 4484 and
FC 4485).
6. Click Configure to return to the Configuration view.
7. I/O Drawer 1 must remain in RIO mode. It cannot be migrated to RIO-2,
because it contains a SCSI adapter and two ESCON adapters which are not
supported by the RIO-2 planar.
8. Right-click on the I/O Drawer 2, and select MES. The Drawer wizard windows
pops-up.
9. Switch to the Storage tab, check the Prefer Dual Loop checkbox, reset the
RIO planar (FC 6563) to zero, and select Two RIO-2 planar (FC 6571), and
click Configure.
10.Repeat Step 9 with I/O drawers 3 and 4.
128
IBM ^ pSeries 670 and pSeries 690 System Handbook
11.In the Wizards window, right-click on the I/O drawer 4, and select duplicate.
There are two entries for I/O drawer 4: one with a 1 on its left, one with a 0.
Select the one preceded by a 1, which corresponds to the updated drawer.
Figure 3-29 Duplicating I/O drawer 4
12.The I/O drawer wizard presents the configuration for I/O drawer 5. Click
Configure to create a new drawer identical to drawer 4.
13.If you display the message view in the Configuration area, you may notice that
e-config has configured features conversions for the replacement of the MCM
and Memory cards.
Chapter 3. Using the IBM Configurator for e-business
129
Figure 3-30 Messages view during feature upgrade
14.The message window also contains errors that need to be fixed before you
can generate a configuration report.
15.Right-click the error:
Rack 1 requires MES/Configure to adjust the number of bulk power
regulators.
Select MES and jump to Wizard, and click Configure. e-config adjusts the
number of Bulk Power features and the number of IBF features. The related
error message disappears from the message view.
16.You can now perform the final validation and generate the configuration
reports.
17.If you select the Proposal view, and scroll down to the list of replaced,
removed and added features, you may notice that e-config has taken care of
the L3 cache, Pass-though modules, Bulk power and cables features of the
CEC and I/O drawers (see lines in bold in Example 3-1).
130
IBM ^ pSeries 670 and pSeries 690 System Handbook
Example 3-1 Resulting configuration changes for CEC features
Removals
5257 PROC.BUS PASS THROUGH MOD.
6404 SUPP.PROC.REM. I/O LOOP ATT.,2
6410 REMOTE I/O LOOP AD,4 LOOP
Additions
4485
5246
6184
6189
6202
6418
6419
4199
16GB Memory Card, 567MHz, Outward Facing
8-Way POWER4+ Turbo Processor Option
POWER C.GROUP,4TH PROC.MOD.
DC POWER CONV.,CEC,ADD
Power Cable Group, CEC to Power Controller,
Support Processor with Remote I/O-2 (RIO-2)
Support Processor with Remote I/O-2 (RIO-2)
128MB Level 3 Cache (4 x 32MB), 567MHz
-1
-1
-1
2
1
1
1
1
1
2
1
............
Removals
3149 REMOTE I/O CABLE, 2 M
6563 I/O DRAWER PCI PLANAR,10 SLOT
Additions
3156 RIO-2 (Remote I/O-2) Cable, 1.75M
6571 I/O Drawer PCI-X Planar, 10 Slot, 2 Integrat
-2
-2
4
2
Chapter 3. Using the IBM Configurator for e-business
131
132
IBM ^ pSeries 670 and pSeries 690 System Handbook
4
Chapter 4.
Capacity Upgrade on
Demand
The pSeries 670 and pSeries 690 systems can be shipped with non-activated
resources (processors and/or memory), which may be purchased and activated
at a future point in time without affecting normal machine operation. This ability is
called Capacity Upgrade on Demand (CUoD) and provides flexibility and
improved granularity in processor and memory upgrades.
This chapter includes the following topics:
򐂰
򐂰
򐂰
򐂰
򐂰
Section 4.1, “What’s new in CUoD” on page 134
Section 4.2, “Description of CUoD” on page 134
Section 4.3, “Activating CUoD resources” on page 146
Section 4.4, “Dynamic Processor Sparing” on page 154
Section 4.5, “On/Off Capacity on Demand” on page 155
Note: This chapter covers the CUoD functions available as of May 2003.
Enhancements and changes may appear in the future.
© Copyright IBM Corp. 2003. All rights reserved.
133
4.1 What’s new in CUoD
With the May 2003 announcement, there are several new CUoD enhancements
to the CUoD systems:
򐂰 In addition to the processor CUoD capability for the 1.1GHz pSeries 670, 1.1
and 1.3 GHz pSeries 690 — the processor CUoD capability for new 1.5 GHz
pSeries 670, 1.5 and 1.7 GHz pSeries 690 is introduced and described in
4.2.3, “Supported CUoD Processor configurations” on page 138.
򐂰 Memory CUoD capability is introduced to all the pSeries 670 and pSeries 690
systems, providing flexibility and fine-granularity for memory for future growth.
The supported configuration is described in 4.2.4, “Supported CUoD Memory
configurations” on page 141.
򐂰 To improve the response time to unpredictable increased workload, the Trial
CoD feature is introduced to the CUoD systems for processors and memory.
This feature allows immediate activation of the non-activated resources on
CUoD systems, which is described in 4.2.1, “Trial CoD function” on page 135.
Section 4.3.2, “Trial CoD processor and memory” on page 151 also describes
how to use this feature from the HMC.
򐂰 To cater for periodical peak workload, such as monthly batch jobs, which the
systems are likely required for additional resources for a short period, there is
a new feature called On/Off Capacity on Demand (On/Off CoD), which is
described in 4.5, “On/Off Capacity on Demand” on page 155.
Attention: On/Off Capacity on Demand for pSeries is not announced as an
orderable product when this redbook was published. The contents in 4.5,
“On/Off Capacity on Demand” on page 155 are subject to change.
4.2 Description of CUoD
This section describes the functionality of Capacity Upgrade on Demand (CUoD)
for processors and memory on the pSeries 670 and pSeries 690 systems. CUoD
systems can be configured with non-activated resources (processors and/or
memory). These non-activated resources can be enabled dynamically and
non-disruptively to the systems without the need to schedule downtime for the
engineer to install additional resources through purchased activation codes at a
certain point in time. It provides flexibility and improved granularity in processor
upgrades and is very useful to anticipate future growth.
The design is based on the use of a CUoD capacity card that has a smart chip to
provide the required high level of asset protection for the CUoD function. This
capacity card is placed in a slot in the primary RIO book (FC 6404) or primary
134
IBM ^ pSeries 670 and pSeries 690 System Handbook
RIO-2 book (FC 6418). This CUoD capacity card tracks activated CUoD
resources.
A new feature called Trial CoD is introduced to provide immediate activation of
the CUoD resources temporarily 30 days without the need for activation code.
This is to allow customers to use the resources while they are ordering the
activation codes from IBM. The additional temporary resources can either be
configured as permanent resources later with the activation code or returned
back to the system as CUoD resources.
We also describe a feature associated with CUoD, called Dynamic Processor
Sparing in 4.4, “Dynamic Processor Sparing” on page 154. It provides protection
against a possible processor failure, by activating a processor from the pool of
non-activated CUoD processors.
The information at the following URL briefly explains the CuOD process:
http://www.ibm.com/servers/eserver/pseries/cuod/
4.2.1 Trial CoD function
Trial CoD is a new feature for CUoD systems. With other CUoD offerings,
customers with CUoD systems must purchase the activation codes from IBM
before the non-activated CUoD resources can be activated to meet the increase
in workload. With this Trial CoD feature, customers can now activate the required
non-activated CUoD resources immediately and after that, proceed to purchase
those resources from IBM.
The following basic rules apply for a pSeries 670 or pSeries 690 system with Trial
CoD feature:
򐂰 After the CUoD resources are activated through Trial CoD function, a
customer must either buy part or all of the activated CUoD resources from
IBM as described in 4.3.1, “CUoD resources activation and order process” on
page 147 or return the activated CUoD resources back to the system within
30 days.
򐂰 The Trial CoD function for processors and memory are activated separately.
򐂰 The customer is allowed one use of Trial CoD:
– When the system first boots up, or
– After the customer has purchased the activated CUoD resources which
were previously activated through Trial CoD function
There are several advantages of using Trial CoD:
򐂰 It improves the response time to meet unpredictable increase in workload.
Chapter 4. Capacity Upgrade on Demand
135
򐂰 A customer can evaluate the performance of the system after activating the
CUoD resources before placing the order for the activation codes.
4.2.2 Overview of CUoD configurations
The following basic rules apply for a pSeries 670 or pSeries 690 system with the
CUoD feature:
򐂰 A CUoD system requires a CUoD capacity card to be placed in a slot in the
primary RIO book (FC 6404) or primary RIO-2 book (FC 6418) for 1.1 and 1.3
GHz systems.
򐂰 A CUoD system requires a CUoD capacity card to be placed in a slot in the
primary RIO-2 book (FC 6418) for 1.5 and 1.7 GHz systems.
򐂰 A system without CUoD resources does not require a CUoD capacity card.
򐂰 For Full System Partition systems, activation of processors and/or memory
will require a reboot.
򐂰 MCMs can be either fully activated MCMs or four out of eight-way active
CUoD MCMs (see 4.2.6, “Logical and physical entities” on page 143 for
further information about how the system activates processors). Additional
two-way activations may be purchased from IBM and activated on these
CUoD MCMs by entering a activation key on the HMC.
򐂰 Affinity partitioning is not supported on CUoD systems.
򐂰 CUoD resources are not supported on Linux-only systems at this point.
򐂰 CUoD is not supported on HPC systems.
򐂰 On partitions that support DLPAR, activation of CUoD resources is supported
without the need for a reboot of the partitions.
򐂰 A CUoD capacity card is required to be present and functional to boot the
system if any CUoD processors are present in the system. If a CUoD memory
resource is present without a CUoD capacity card, the system will
deconfigure the CUoD memory card at boot time, but will allow the system to
boot if there is non-CUoD memory cards.
򐂰 A CUoD capacity card from another system will allow the system to be
rebooted, but none of the CUoD functions (screens, activations) can be used.
An error message will be sent to the HMC to indicate that the system CUoD
functions and supporting screens are not available.
򐂰 CUoD resources can be moved from one CUoD system to another CUoD
system; however, activation entitlement stays with the serial number of the
purchase.
򐂰 Dynamic Processor Sparing is available only on systems running on dynamic
logical partitioning (DLPAR).
136
IBM ^ pSeries 670 and pSeries 690 System Handbook
򐂰 To activate CUoD resources, a key called the activation code has to be
entered on the HMC. The maximum number of invalid activation code entries
is five. If more than five activation codes are entered, the system must be
rebooted before another activation code can be entered. Some types of
failures are not counted against the five invalid attempts, including:
– User mis-typed the activation code. (There are unique features in the
activation codes that are checked.)
– Previous activation code for the same system.
– Activation code from a different system that contains a different processor
(speed) type.
– Activation code for memory from a different system is entered and this
system does not contain any CUoD memory cards.
򐂰 For activation codes for CUoD processors, the following microcode levels,
HMC software levels and operating system versions (with fixes) are required
(including Full System Partition):
Note: This applies only for the POWER4 1.1 and 1.3 GHz processor
systems.
– 7040 - 671/681 at 10/2002 system microcode update or later
– HMC for pSeries with V1.3 or later
– AIX 5L Version 5.1 with 5100-03 Recommended Maintenance Level with
APAR IY36013 or later
Note: The AIX 5L Version 5.1 partition (including Full System Partition)
would have to be rebooted to take advantage of the additional
processors.
– AIX 5L Version 5.2
򐂰 For Trial CoD of CUoD processors, the following microcode levels, HMC
software levels and operating system versions (with fixes) are required
(including Full System Partition):
– 7040 - 671/681 at 05/2003 system microcode update or later
– HMC for pSeries with V1.3.2 or later
– AIX 5L Version 5.1 with 5100-04 Recommended Maintenance Level with
APAR IY39795 and APAR IY36771 or later
– AIX 5L Version 5.2 with 5200-01 Recommended Maintenance Level with
APAR IY39795 and APAR IY36772 or later
Chapter 4. Capacity Upgrade on Demand
137
򐂰 For activation codes for CUoD memory and Trial CoD of CUoD memory, the
following microcode levels, HMC levels and operating system versions (with
fixes) are supported (including Full System Partition):
– 7040-671/681 at 05/2003 system microcode update or later
– HMC for pSeries with V1.3.2 or later
– AIX 5L Version 5.2 with 5200-01 Recommended Maintenance Level with
APAR IY39795 and APAR IY36772 or later
Note: For HMC for pSeries with V1.3.2, the first number represents version,
the second number is the release, and the third number is modification.
Important: The small RMO option must be set to boot AIX 5L Version 5.2
when you create an LPAR from the HMC.
4.2.3 Supported CUoD Processor configurations
Processors are activated in pairs because of the dual core technology of
POWER4 and POWER4+ processors and to optimize system performance and
maintain consistent, predictable performance. A complex algorithm in the system
firmware, based on such things as memory groups, available inactive
processors, and possible processor failure or deallocation situations, determines
which pair of processors will be activated. In normal circumstances, both
processor cores in a single chip will be activated. But there are circumstances in
which that may not be possible.
Note: This activation of pairs of processors is also true for On/Off CoD and
Trial CoD.
Table 4-1 on page 138 is an overview of all CUoD processor feature codes for
the pSeries 670 and pSeries 690.
Table 4-1 CUoD processor feature codes
FC
Description
pSeries 670
138
7402
1.1 GHz 4/8-way CUoD MCM + 128 MB (4x 32) L3 cache
7412
1.1 GHz CUoD 2-way Activation for FC 7402
7400
1.5 GHz 4/8-way CUoD MCM + 128 MB (4x32) L3 cache
IBM ^ pSeries 670 and pSeries 690 System Handbook
FC
Description
7410
1.5 GHz CUoD 2-way Activation for FC 7400
pSeries 690
7403
1.1 GHz 4/8-way CUoD MCM
7413
1.1 GHz CUoD 2-way Activation for FC 7403
7404
1.5 GHz 4/8-way CUoD MCM
7414
1.5 GHz CUoD 2-way Activation for FC 7404
pSeries 690 Turbo
7405
1.3 GHz 4/8-way CUoD MCM
7415
1.3 GHz CUoD 2-way Activation for FC 7405
7406
1.7 GHz 4/8-way CUoD MCM
7416
1.7 GHz CUoD 2-way Activation for FC 7406
Table 4-2 shows possible CUoD processor combinations for 1.1 and 1.3 GHz
systems.
Under the MCM columns:
򐂰 The numbers that are not emphasized (for example, 4) indicates the number
of processors that are installed and activated on the MCM.
򐂰 The numbers that are emphasized (for example, 2) indicates the number of
processors that are installed, but can either be activated upon initial
installation of the system or activated through the CUoD activation process on
the MCM in future.
Table 4-2 Supported 1.1 and 1.3 GHz CUoD processor combinations
Model &
System
CUoD
Range
MCM 0
MCM 2
MCM 1
pSeries
670
16-way
12 to 16
8
4/
2/2
pSeries
690
16-way
12 to 16
8
4/
2/2
pSeries
690
24-way
16 to 24
8
4/
2/2
4/
2/2
20 to 24
8
8
4/
2/2
MCM 3
Chapter 4. Capacity Upgrade on Demand
139
Model &
System
CUoD
Range
MCM 0
pSeries
690
32-way
20 to 32
8
24 to 32
28 to 32
MCM 2
MCM 1
MCM 3
2/2
4/
2/2
4/
2/2
8
8
4/
2/2
4/
2/2
8
8
8
4/
2/2
4/
Table 4-3 shows possible CUoD processor combinations for 1.5 and 1.7 GHz
systems.
Table 4-3 Supported 1.5 and 1.7 GHz CUoD processor combinations
Model &
System
CUoD Range
MCM 0
MCM 2
pSeries
670
16-way
8 to 16
4/2/2
4/
2/2
12 to 16
8
4/
2/2
8 to 16
4/2/2
4/
2/2
12 to 16
8
4/
2/2
12 to 24
4/2/2
4/
2/2
4/2/2
16 to 24
8
4/
2/2
4/2/2
20 to 24
8
8
4/2/2
16 to 32
4/2/2
4/
2/2
4/2/2
4/
2/2
20 to 32
8
4/
2/2
4/2/2
4/
2/2
24 to 32
8
8
4/2/2
4/
2/2
28 to 32
8
8
8
4/
2/2
pSeries
690
16-way
pSeries
690
24-way
pSeries
690
32-way
MCM 1
MCM 3
Example 1
We configure a 1.1 GHz 12-way pSeries 670 with twelve processors by
configuring one FC 5256 and one FC 7402. This leaves room to order two FC
7412s to reach the maximum processor configuration.
Example 2
We configure a 1.7 GHz 16-way pSeries 690 with sixteen CUoD processors by
configuring four FC 7406 (4 out of 8-way operational MCM). This leaves room to
140
IBM ^ pSeries 670 and pSeries 690 System Handbook
order eight FC 7416 (2-way processor activation) to reach the maximum 32-way
configuration.
Note: The two-way CUoD activations may also be configured in the initial
order to form additional configurations. For example, a 26 out of 32-way 1.1
GHz pSeries 690 can be configured by selecting one FC 5242, one FC 5252,
two FC 7403s, and one FC 7413.
4.2.4 Supported CUoD Memory configurations
Table 4-4 is an overview of all CUoD memory feature codes for the pSeries 670
and pSeries 690. There are some basic rules for these CUoD memory
combinations:
򐂰 All CUoD memory card must be in identical pairs, that is, 2 x FC 7050 or 2 x
7051.
򐂰 All CUoD memory activation must be in identical pairs, that is, 2 x FC 7060 or
2 x FC 7061.
򐂰 Configurations consisting of one CUoD memory card and one non-CUoD
memory card are not supported.
Table 4-4 CUoD memory feature codes
FC
Description
pSeries 670
7050
16 GB 567 MHz CUoD memory card, inward facing, 8 GB active
7054
32 GB 567 MHz CUoD memory card, inward facing, 16 GB active
7060
4 GB CUoD activation for FC 7050 and FC 7051
7061
4 GB CUoD activation for FC 7054 and FC 7055
pSeries 690
7050
16 GB 567 MHz CUoD memory card, inward facing, 8 GB active
7051
16 GB 567 MHz CUoD memory card, outward facing, 8 GB active
7054
32 GB 567 MHz CUoD memory card, inward facing, 16 GB active
7055
32 GB 567 MHz CUoD memory card, outward facing, 16 GB active
7060
4 GB CUoD activation for FC 7050 and FC 7051
7061
4 GB CUoD activation for FC 7054 and FC 7055
Chapter 4. Capacity Upgrade on Demand
141
The conventional memory configuration table in 2.3.2, “Memory subsystem for
pSeries 690” on page 29 applies to CUoD memory cards. Utilize maximum card
capacity when applying the configuration rules to CUoD memory cards:
򐂰 Partially activated 16 GB CUoD memory cards apply 16 GB configuration
rules.
򐂰 Partially activated 32 GB CUoD memory cards apply 32 GB configuration
rules.
Example 1
We configure a two MCMs pSeries 670 with 32 GB memory by configuring four
FC 7050. This leaves room to order eight FC 7060 to reach the maximum
memory configuration of 64 GB.
Example 2
We configure a four MCMs pSeries 690 with 160 GB memory by configuring four
FC 7054 (16 of 32 GB active inward memory), two FC 7055 (16 of 32 GB active
outward memory) and two FC 4487. This leaves room to order sixteen FC 7060
and eight FC 7061 to reach the maximum memory configuration of 256 GB.
Note: The 4 GB CUoD activations may also be configured in the initial order to
form additional configurations. For example, a two MCM pSeries 690 with 40
GB out of a maximum of 64 GB can be configured by selecting four FC 7050
and two FC 7060.
4.2.5 CUoD resource sequencing
Once processors and memory from the CUoD resources are activated, an
algorithm is used to enable the physical resources in a specific order to ensure
the following characteristics:
򐂰 Activations are not tied to actual physical CUoD resources.
򐂰 All processor activations will be in full chip increments (this means two
processors at a time on the same chip).
򐂰 Hot sparing may cause performance imbalance within the system since two
chips will have only a single core used.
򐂰 Processor will be used based upon best performance tuning rather than
physical relationship to CUoD resources. The entitled resources used will be
selected independently of whether the resource is from a CUoD component
or a non-CUoD component.
142
IBM ^ pSeries 670 and pSeries 690 System Handbook
4.2.6 Logical and physical entities
From an ordering point of view it might seem that a base MCM is always fully
operational and a CUoD MCM is initially always four out of eight-way operational.
This is the logical view of CUoD. However, whichever specific physical entities
will be activated, CUoD is completely determined by the sequencing rules. Since
these rules do not consider the type of the resource (a base resource or CUoD
resource), there will be situations where specific base resources are not
activated, and specific CUoD resources are activated though not purchased.
Also, the technical implementation of CUoD in firmware is implemented such that
all theoretical configurations are covered (not only the recommend balanced
memory and L3 cache configurations, and dual processor activations). Secondly,
the technical implementation is not only meant for the pSeries 670 or pSeries
690, but also for future pSeries machines, possibly with finer CUoD granularity.
On actual shipping of pSeries 670 and pSeries 690 systems, memory guidelines
and supported configurations (as described in this book and in Table 2-6 on
page 44) are followed, therefore, only balanced memory configurations will exist
and actual CUoD activations and sequencing rules will be a subset of those
mentioned in 4.2.5, “CUoD resource sequencing” on page 142.
Let us clarify these rules using the following two examples:
In both cases we consider a 16 out 24-way operational pSeries 690 configuration
(three MCMs: one base MCM and two CUoD MCMs). We assume that the
memory configuration is balanced for all MCMs, as this is recommended for
practically all configurations.
Example 1
Total system memory is 40 GB: 8+8, 8+8, and 4+4 for MCM 0, 2, and 1,
respectively. Due to sequencing of large balanced memory groups, all
processors on MCM 0, and subsequently on MCM 2, are active, and none are
active on MCM 1. The active processors will see all the memory, including the
memory attached to MCM 1, totalling 40 GB.
Example 2
Total system memory is 16 GB: 4+4, 4+4, and 0+0 for MCM 0, 2, and 1,
respectively. Again all processors on MCM 0 and 2 will be active. Now, if an
activation code for two CUoD processors is entered, they will become active on
MCM 1, which has no memory attached. All 18 processors will use the 16 GB
memory of MCM 0 and 2.
Chapter 4. Capacity Upgrade on Demand
143
4.2.7 CUoD license screen
Upon the first boot of the pSeries 670 or pSeries 690 CUoD systems, the “Click
to Accept” message is presented to the system administrator, and it should be
accepted for the initial boot of this system. The “Click to Accept” message will be
shown on the HMC graphical user interface as shown in Figure 4-1 on page 144.
Figure 4-1 Click to Accept on the HMC at initial boot1
The CUoD License function works as explained in the following:
1. On every boot, system firmware looks for a Click to Accept (CTA) accepted bit
stored in the CUoD capacity card that indicates that the CUoD Click To
Accept screen has displayed and the license conditions accepted. If the bit is
set, the system will continue the boot. If the CTA accepted bit is not set,
A100C2AC will be displayed in the operator panel, and firmware will contact the
HMC and request that the “Click to Accept” screen be displayed. Firmware
will then wait for the HMC to return with the status that the Click To Accept
license conditions have been accepted. Once accepted the system will
continue the boot.
2. There are two possible responses to accept in the “Capacity Upgrade on
Demand License Agreement” window (see Figure 4-2):
1
The host name of the HMC is hidden in the title bar.
144
IBM ^ pSeries 670 and pSeries 690 System Handbook
– If Do not show this information again is selected and accepted, the boot
will continue and the HMC will not display the CUoD Click To Accept
screen again.
– If Do not show this information again is not selected and accepted, the
boot will continue, but the CUoD Click To Accept screen will be displayed
on the HMC again on the next reboot.
Figure 4-2 Click to Accept window
Note: If you click the Do not accept button, the boot process will not
continue.
3. At boot time, the system firmware checks if the system power source (wall
plug) has been removed as part of a controlled shutdown prior to the boot. If
power had been removed, system firmware will present the “Click To Accept”
screen on the next boot. This will cover the case that the system may have
been sold or transferred and is now being booted by a new customer. The
“Click To Accept” screen will not be displayed again if power had been
removed due to a power outage.
A service processor menu to reset the CUoD Click To Accept settings is also
implemented.
4.2.8 CUoD error messages
The system firmware will detect some error conditions concerning CUoD. The
HMC will indicate several errors (as listed in Table 4-5) to the operator in case of
CUoD failure situations.
Chapter 4. Capacity Upgrade on Demand
145
When a user receives out-of-compliance messages from a CUoD system, he can
either shutdown one or more partitions containing sufficient processors/memory
or to return the exact amount of processors/memory by using dynamic logical
partitioning to get the system back in compliance.
Table 4-5 CUoD error codes and messages
Error code
Description
1
Processor capacity activation failed.
2
Memory capacity activation failed.
3
Processor capacity activation code is not valid.
4
Memory capacity activation code is not valid.
5
Processor capacity activation code is not valid, next failed attempt will
require a re-boot.
6
Memory capacity activation code is not valid, next failed attempt will
require a re-boot.
7
Processor capacity activation code is not valid, a re-boot is required to
make another attempt.
8
Memory capacity activation code is not valid, a re-boot is required to
make another attempt.
9
Activation Code entered incorrectly, key check error. This is most likely a
keying error and the user would be instructed to please re-enter the
Activation Code.
10
CUoD capacity card detected from another system or has corrupted
data. Contact service provider to replace the card.
11
Request not allowed - Trial CoD function selected had been used.
12
Request not allowed - CUoD capacity card is failing or has corrupted
data. Contact your service provider to replace the card.
13
Request not allowed - More processors were selected than are available
for immediate activation.
14
Request not allowed - More memory was selected than is available for
immediate activation.
4.3 Activating CUoD resources
Here we describe these activation processes on a pSeries 670 or pSeries 690
CUoD system:
146
IBM ^ pSeries 670 and pSeries 690 System Handbook
򐂰 4.3.1, “CUoD resources activation and order process” on page 147
򐂰 4.3.2, “Trial CoD processor and memory” on page 151
4.3.1 CUoD resources activation and order process
To utilize the CUoD resources (processors and memory) on a pSeries 670 or
pSeries 690 CUoD machine permanently, the following steps are required:
1. A customer gathers system-specific data from the HMC, as follows:
a. When you have a CUoD-managed system, the Capacity Upgrade on
Demand menu will appear on the HMC when you right-click the managed
system as shown in Figure 4-3.
Figure 4-3 CUoD menu on the HMC
b. On the managed system, select Capacity Upgrade on Demand ->
Processor or Memory -> Processor Capacity Settings or Memory
Capacity Settings and a menu displaying the resources on the system
will appear (see Figure 4-4).
Chapter 4. Capacity Upgrade on Demand
147
Figure 4-4 Processor Capacity Settings
c. Select Display Order Information and the Save Processor or Memory
Order Information will appear (see Figure 4-5).
Figure 4-5 Save Processor Order Information
148
IBM ^ pSeries 670 and pSeries 690 System Handbook
d. Save the processor or memory order information either to a remote
system or a diskette.
2. This information can be sent to an IBM Sales Representative using this URL
(Figure 4-6):
http://www.ibm.com/servers/eserver/pseries/cuod/vpd_form.html
Figure 4-6 Send CUoD and system data
Chapter 4. Capacity Upgrade on Demand
149
Or, this information can be sent using Service Agent to send CUoD and
system data electronically to the IBM Machine Reported Product Data
(MRPD) database. Service Agent is the preferred method to gather the data
and send it to IBM.
3. The customer places an MES/Upgrade order to a IBM sales representative
who will then places the CUoD MES/Upgrade for activation via the e-config.
This order process will produce a billing.
4. The CUoD application continually looks for CUoD orders, and when an order
comes in it will look for CUoD data input either from the pSeries CUoD URL or
from the MRPD database, and will put system CUoD data into the CUoD
application. The CUoD application will request the CUoD activation code as
ordered from the CUoD activation code tool. The CUoD activation code tool
will return the CUoD activation code to the CUoD application.
Note: Step 4 is an IBM internal process.
5. The generated CUoD activation code, a unique code with a system of 34
characters, will be posted to a Web site where the customer can access, and
it will also be printed by IBM manufacturing and sent by conventional mail to
the customer.
6. When the customer receives the activation code, he or she needs to do the
following steps on the HMC to activate the resources:
a. On the CUoD system that you want to activate the CUoD resources,
right-click the managed system and select Capacity Upgrade on Demand
-> Activate, and a menu will appear (see Figure 4-7).
Figure 4-7 Activate CUoD Resources
b. Enter the CUoD activation code.
7. The activation code is passed to the service processor for authentication. This
includes error checking on the key check value and sequence number,
decryption of the data, and further error checking to determine that the
activation code is for this system.
150
IBM ^ pSeries 670 and pSeries 690 System Handbook
8. The service processor stores the data in the CUoD capacity card.
9. When the system is running the Full System Partition, a reboot is required to
activate the additional resources.
10.When the system is running a partitioned environment, the service processor
notifies the system firmware and HMC about the change in licensed status of
the additional resources. The customer, using existing HMC screens, decides
how to allocate the additional resources to partitions. The normal dynamic
logical partitioning (DLPAR) procedure accomplishes the CUoD task as a
sequence of basic resource add or remove operations. For partitions that are
not running dynamic logical partitioning (DLAPR), allocation of additional
resources will require a reboot of the partitions.
The following is a list of the pSeries CUoD data that will be obtained from the
HMC and should be faxed by the IBM representative to the CUoD administrator
or is electronically sent to the MRPD database for input to the CUoD activation
code generation tool:
򐂰 System type (four ASCII characters)
򐂰 System serial number (eight ASCII characters: pp-sssss)
򐂰 CUoD capacity card CCIN (four ASCII characters)
򐂰 CUoD capacity card serial number (10 ASCII characters: pp-sssssss)
򐂰 CUoD capacity card unique ID (16 ASCII characters)
򐂰 CUoD resource identifier (four ASCII characters)
򐂰 Activated CUoD function: CUoD increments in use, resources currently paid
for (four ASCII characters)
򐂰 CUoD sequence number (four hexadecimal characters)
򐂰 CUoD activation code entry check (1 byte hex check sum, two ASCII
characters—based on items: 1, 2, 3, 4, 5, 6, 7, and 8)
This information is required to generate the activation code (see Figure 4-6 on
page 149).
4.3.2 Trial CoD processor and memory
In this section, we describe the necessary steps to activate the CUoD resources
(processors and memory) using Trial CoD function on the HMC:
1. On the HMC, right-click the managed system and select Capacity Upgrade on
Demand -> Processor -> Processor Activate Immediate for processors, or
select Capacity Upgrade on Demand -> Memory -> Memory Activate
Immediate for memory as shown in Figure 4-8.
Chapter 4. Capacity Upgrade on Demand
151
Figure 4-8 To activate the resources using Trial CoD function
2. A menu will appear as shown in Figure 4-9. For processor activation, you can
add a minimum of 1 processor and up to the maximum of non-activated
CUoD processors available in the system. For memory activation, you can
add a minimum of 1 GB and up to a maximum of non-activated CUoD
memory available in the system.
Figure 4-9 CUoD Trial CoD Processor Capacity (Next) screen
152
IBM ^ pSeries 670 and pSeries 690 System Handbook
Important: All of the screen images for the Trial CoD function were taken
before the official product announcement. In the window titles or menus, any
text that appears as "Activate Immediate" should be read as "Trial CoD". It will
be modified in the future software maintenance update of the HMC software
product.
3. After entering the required amount of resources, click Next and a summary of
the activated and non-activated resources will be displayed (see Figure 4-10).
Figure 4-10 CUoD Trial CoD Processor capacity (Finish) screen
4. By clicking Finish, the customer has agreed to either purchase part or all of
the activated resources or return them back to the system by disabling the
Trial CoD function within 30 days.
5. If the customer chooses to purchase the resources, he or she needs to
execute the steps as described in 4.3.1, “CUoD resources activation and
order process” on page 147.
Chapter 4. Capacity Upgrade on Demand
153
6. If the customer decides to stop using the resources, he or she will need to
disable the Trial CoD function to return the resources back to the system. On
the HMC, select Capacity Upgrade on Demand -> Processor -> Processor
Disable Activate Immediate for processors, or select Capacity Upgrade on
Demand -> Memory -> Memory Disable Activate Immediate for memory,
and a warning message will appear as shown in Figure 4-11.
Figure 4-11 Warning message to Disable Processor Trial CoD Capacity
7. Click Yes to confirm the customer does not require the resources anymore
and the resources will be returned to the system as non-activated CUoD
resources.
Important: The customer will not be able to use Trial CoD function until the
customer has purchased some of the CUoD resources which were activated
through Trial CoD function previously.
4.4 Dynamic Processor Sparing
When you have a CUoD system (that is, a system with at least one CUoD MCM),
a feature called Dynamic Processor Sparing is automatically provided with it.
Dynamic Processor Sparing is the capability of the system to disable a failing
processor and enable a non activated CUoD processor without any manual
intervation. Non-activated CUoD processors are processors that are physically
installed in the system, but not yet activated. They cannot run jobs or tasks
unless they become activated via a CUoD processor activation code.
Note: If all the CUoD processors in the system are activated, even if these
processors are not allocated to any partition, Dynamic Processor Sparing will
not be enabled.
The processor that will be used to replace a failing processor will be the first
available CUoD non-activated processor in the CUoD resource sequence, as
described in 4.2.5, “CUoD resource sequencing” on page 142.
154
IBM ^ pSeries 670 and pSeries 690 System Handbook
Both base processors and licensed CUoD processors may be replaced with
unlicensed CUoD processors for Dynamic Processor Sparing when failing.
When Dynamic Processor Sparing is performed, the system will be out of CUoD
compliance for a period of time between when the spare processor is allocated
and the failed processor is deallocated. By design, AIX must request the spare
processor, and in doing so AIX is committing to returning the failing processor in
a timely fashion. If it fails to return the processor, the CUoD compliance
enforcement code will detect the out of compliance condition. System firmware
will ignore anything out of compliance during hot sparing for a sufficient time
period for AIX to return the failing processor. The out of compliance condition will
occur when more processors are active than what the customer is entitled to run
(the number of base processors plus the number of CUoD processors by the
customer).
Notes:
򐂰 Dynamic Processor Sparing requires AIX 5L Version 5.2 or higher, and the
system must be run in a partitioned environment. The Dynamic Processor
Sparing requires that the CPU guard attribute is set to enable (this is the
default value on AIX 5L Version 5.2 or higher), as explained in “Enabling
CPU Guard” on page 170.
򐂰 There is no “Dynamic Memory Sparing” function. Non activated memory (if
it is available in the system) will be sequenced in on the next full system
re-boot if there is any faulty memory detected.
4.5 On/Off Capacity on Demand
This section describes the functionality of On/Off Capacity on Demand (On/Off
CoD) for processors on the pSeries 670 and pSeries 690 systems with CUoD
resources. With On/Off CoD, a customer can enable and disable the
non-activated processors dynamically and non-disruptively to the systems
depending on their workload at a certain point in time. It provides flexibility to
provide additional temporary processing power to the system during peak
workload (for example, month-end reporting batch jobs).
The following are the basic features for a pSeries 670 or pSeries 690 system with
the On/Off CoD:
򐂰 Each On/Off CoD activation entitles the customer to 60 processor days (PDs),
defaulting to two processors for 30 days. Processor days (PDs) are actual in
use days and not calendar days.
򐂰 Multiple On/Off CoD activation features are allowed, each one entitling the
customer to an additional 60 PDs from the time it is entered into the system.
Chapter 4. Capacity Upgrade on Demand
155
򐂰 A new customer interface will be introduced for the customer to manage the
On/Off CoD processors.
򐂰 If the customer deactivates the On/Off CoD-activated processors, the
remaining PDs are “banked” for use in the future.
Example 1
When a customer entered two On/Off CoD activation codes. Each activation
code by default gives him access to two processors for 30 days, or 60 PDs, in this
case a total of 120 PDs. If four non-activated processors are available in the
system, there are two possible configurations:
򐂰 Activate 4 processors for 30 days, or
򐂰 Activate 2 processors for 60 days
156
IBM ^ pSeries 670 and pSeries 690 System Handbook
5
Chapter 5.
Reliability, availability, and
serviceability
The terms reliability, availability, and serviceability (RAS) are widely used
throughout the computer industry as an indication of a product’s failure
characteristics. RAS refers to a collection of interdependent product attributes
that, when taken together, attempt to measure how often a product fails, how
quickly it can be repaired, and the overall system disruption caused by a failure.
This chapter describes the various features and mechanisms that are
implemented in the pSeries 670 and pSeries 690 servers to minimize failures,
and isolate and recover them if possible, by providing the following topics:
򐂰
򐂰
򐂰
򐂰
򐂰
򐂰
򐂰
Section 5.1, “What’s new in serviceability” on page 158
Section 5.2, “RAS features” on page 158
Section 5.3, “Predictive functions” on page 159
Section 5.4, “Redundancy in components” on page 163
Section 5.5, “Fault recovery” on page 166
Section 5.6, “Serviceability features” on page 177
Section 5.7, “AIX RAS features” on page 206
All this is developed to keep the impact on server operations for the customer as
small as possible.
© Copyright IBM Corp. 2003. All rights reserved.
157
5.1 What’s new in serviceability
All the RAS features that are incorporated in the previous pSeries 670 and
pSeries 690 remained with the May 2003 announcement. A new feature to
provide the customer the ability to survey and update the microcode levels was
introduced. This will help customers ensure that their systems’ microcode levels
are up-to-date. Three sections have been added to explain the serviceability
features:
򐂰 Section 5.6.1, “Back up of HMC” on page 180 provides an overview of two
types of back up available in the HMC and the usage.
򐂰 Section 5.6.2, “Upgrading HMC” on page 181 describes the steps involved in
upgrading the HMC to the level that the microcode updates feature can be
used.
򐂰 Section 5.6.3, “Microcode Updates function” on page 184 describes the steps
to survey and update the microcode from the HMC.
5.2 RAS features
Both the pSeries 670 and pSeries 690 bring new levels of availability and
serviceability to the enterprise-level UNIX servers. Advanced functions
presented on the previous models were enhanced, and new technologies have
been added, to reduce faults and minimize their impacts. High-quality
components and rigorous verification and testing during the manufacturing
processes contribute to the reduction of failures in the system. The diagnostic
goal of these RAS features is to isolate the Field Replaceable Units (FRU) challot
to 95 percent to a single FRU; of course there will still be a chance that more than
one FRU is failing, but the implemented features in the pSeries 670 and pSeries
690 will keep the customer impact as small as possible.
The pSeries 670 and pSeries 690 RAS design enhancements can be grouped
into four main areas:
򐂰 Predictive functions: These are targeted to monitor the system for possible
failures, and take proactive measures to avoid the failures.
򐂰 Redundancy in components: Duplicate components and data paths to
prevent single points of failure.
򐂰 Fault recovery: Provide mechanisms to dynamically recover from failures,
without system outage. This includes dynamic deallocation of components
and hot-swap capability.
򐂰 Serviceability features: Enable the system to automatically call for support,
and provide tools for rapid identification of problems.
158
IBM ^ pSeries 670 and pSeries 690 System Handbook
The pSeries 670 and pSeries 690 RAS presents features in all these categories,
as described in the following sections.
5.3 Predictive functions
In a mission-critical application, any outage caused by system failure will have an
impact on users or processes. The extent of the impact depends on the type of
the outage and its duration.
Unexpected system outages are the most critical in a system. The disruption
caused by these outages not only interrupts the system execution, but can
potentially cause data problems of either loss or integrity. Moreover, the recovery
procedures after such outages are generally longer than for planned outages,
because they involve error recovery mechanisms and log analysis.
Planned system outages are also critical in a mission-critical environment.
However, the impact can be minimized by adequately planning the outage time
and the procedures to be executed. Applications are properly shut down, and
users are aware of the service stop. Therefore, there is less exposure when
doing planned outages in a system.
Reliability engineering is the science of understanding, predicting, and
eliminating the sources of system failures. Therefore, the ability to detect
imminent failures, and the ability to plan system maintenance in advance are
fundamental to reducing outages in a system, and to effectively implement a
reliable server.
5.3.1 Service processor
The service processor is the interface between the HMC and the pSeries 670
and pSeries 690. It also provides functions for the Service Focal Point
application.
Most of the predictive analysis and error recovery actions are directly related to
or assisted by the service processor. However, there is no dependency between
a server operation and the service processor. This means that even if the service
processor fails, the server continues operating, and the running partitions are not
affected by the failure.
What the service processor is
A service processor is a separate independent processor that provides hardware
initialization during system IPL, operation monitoring of environmental and error
events, and maintenance support for the pSeries 670 and pSeries 690. The
Chapter 5. Reliability, availability, and serviceability
159
service processor communicates with the pSeries 670 and pSeries 690 through
attention signals from the hardware and read/write communication through the
JTAG ports between the service processor and all the pSeries 670 and pSeries
690 chips. This diagnostic capability is simultaneous, asynchronous, and
transparent to any system activity running on the machine.
Figure 5-1 on page 160 illustrates an overview of how the service processor
works.
Configuration
data
Scan access
to checkers
AIX
RTAS
Memory
O
O
O
50
120
40
30
100
C
20
10
Service
Processor
POWER4
POWER4
POWER4
POWER4
L2,L3
L3
L2,
L3
L2,
L2,
L3
caches
caches
caches
caches
Operator Panel
OK
P
C
I
Base
I/O
0
10
20
30
40
50
Environmental
monitoring
F
80
60
40
20
0
20
40
60
Monitor and
control power
I2 C
ISA
Messages to
and from AIX
Environmental
Interrupts
Serial Ports
Local/remote
console
S1
S2
Diskette
Firmware update
NVRAM
Customization
and logging
Battery
clock
11
12
1
10
2
3
9
4
8
7
6
5
Timed
power-on
Figure 5-1 Service processor schematic
The NVRAM has a special function for the service processor. It acts like a
mailbox for the service processor, system firmware, and the AIX running on the
pSeries 670 and pSeries 690. An FRU callout, for example, will be removed by
the system firmware and passed to AIX, then written into the AIX system error
log.
5.3.2 First Failure Data Capture (FFDC)
The pSeries 670 and pSeries 690 have the capability of error detection and
logging, using AIX and service processor functions. These functions work in a
special way to capture all possible failures, and to store information about them.
160
IBM ^ pSeries 670 and pSeries 690 System Handbook
First Failure Data Capture is the keystone to building a system problem
determination strategy around the capture and analysis of errors as they happen,
as opposed to attempting to recreate them later. This not only provides the exact
information about the status of the system when the failure happened, but it also
provides a way to detect and capture information related to temporary failures, or
failures related to server load. Hardware error checkers are strategically placed
on the system, operating full time to detect and capture precise error information.
All error checkers have a blocking logic so that for every detected error the error
is recorded only by the first checker that encounters it. The error information is
stored in Fault Isolation Registers (FIRs) to be later examined.
There are over 5600 fault isolation register bits representing over 15000 internal
error checkers. All error check stations have been placed in the data and control
paths of pSeries 670 and pSeries 690 systems to deterministically isolate
physical faults based on run-time detection of each unique failure that may occur.
Run-time error diagnostics are deterministic in that for every check station, the
unique error domain for that checker is defined and documented.
Figure 5-2 shows a schematic example of error checker positioning and FIR
updating.
Error Checkers
CPU
Fault Isolation Register (FIR)
(the unique fingerprint of
each error captured)
L1 Cache
L2/L3 Cache
Service
Service
Processor
Processor
Log Error
Memory
Nonvolatile
RAM
Disk
Figure 5-2 FFDC error checkers and fault isolation registers
Chapter 5. Reliability, availability, and serviceability
161
5.3.3 Predictive failure analysis
Statistically, there are two main situations where a component has a catastrophic
failure: Shortly after being manufactured, and when it has reached its useful life
period. Between these two regions, the failure rate for a given component is
generally low, and normally gradual. A complete failure usually happens after
some degradation has happened, be it in the form of temporary errors, degraded
performance, or degraded functionality.
The pSeries 670 and pSeries 690 have the ability to monitor critical components
such as processors, memory, cache, I/O subsystem, and internal disks, and
detect possible indications of failures. By continuously monitoring these
components, upon reaching a threshold, the system can isolate and deallocate
the failing component without system outage, thereby avoiding a complete
failure.
5.3.4 Component reliability
The components used in the CEC provide superior levels of reliability that are
available and undergo additional stress testing and screening above and beyond
the industry-standard components that are used in several UNIX-based systems
today.
Fault avoidance is also served by minimizing the total number of components,
and this is inherent in POWER4 chip technology, with two processors per chip. In
addition, internal array soft errors throughout the POWER4 chip are
systematically masked using internal ECC recovery techniques whenever an
error is detected. Going beyond ECC in the memory subsystem, the basic
memory DIMM technology has been significantly improved in reliability through
the use of more reliable soldered connections to the memory card.
5.3.5 Extended system testing and surveillance
The design of the pSeries 670 and pSeries 690 aids in the recognition of
intermittent errors that are either corrected dynamically or reported for further
isolation and repair. Parity checking on the system bus, cyclic redundancy
checking (CRC) on the remote I/O bus, and the use of error correcting code on
memory and processors contribute to outstanding RAS characteristics.
During the boot sequence, built-in self test (BIST) and power-on self test (POST)
routines check the processors, cache, and associated hardware required for a
successful system start. These tests run every time the system is powered on.
162
IBM ^ pSeries 670 and pSeries 690 System Handbook
Additional testing can be selected at boot time to fully verify the system memory
and check the chip interconnect wiring. When a system reboots after a hard
failure, it performs extended mode tests to verify that everything is working
properly and that nothing was compromised by the failure. This behavior can be
overridden by the systems administrator.
5.4 Redundancy in components
The pSeries 690 and pSeries 670 system design incorporates redundancy in
several components to provide fault-tolerant services where failures are not
allowed. Power supplies, fans, blowers, boot disks, I/O links, and power cables
offer redundancy to eliminate single points of failure. Some of these features are
highlighted, as described in the following sections.
5.4.1 Power and cooling
The pSeries 670 and pSeries 690 provide full power and cooling redundancy,
with dual power cords and variable-speed fans and blowers, for both the CEC
and the I/O drawers.
Within the CEC rack, the N+1 power and cooling subsystem provides complete
redundancy in case of failures in the bulk or regulated power supplies, the power
controllers, and the cooling units, as well as the power distribution cables. As on
the zSeries server, concurrent repair is supported on all of the CEC power and
cooling components. See Table 5-1 on page 173 for the hot-pluggable
components.
Note: The pSeries 670 supports either single-phase power or three-phase,
while pSeries 690 supports only three-phase power.
There is also a redundant feature called internal battery features (IBF) designed
to maintain system operation during brown-out conditions. In case of a total
power loss, the batteries can be used to execute a complete shutdown of the
system in an ordered way. For full power loss protection, the pSeries 670 and
pSeries 690 supports optional uninterruptible power supply (UPS) systems in
addition to, or in place of, the IBF features. You should see the IBF feature as a
redundancy feature only; it will not replace the UPS capabilities.
In case of a fan or blower failure, the remaining fans automatically increase
speed to compensate for the air flow from the failed component.
Chapter 5. Reliability, availability, and serviceability
163
5.4.2 Memory redundancy mechanisms
There are several levels of memory protection. From the internal L1 caches to
the main memory, several features are implemented to assure data integrity and
data recovery in case of memory failures.
򐂰 The L1 caches are protected by parity, and if an error is detected, the data is
fetched from the L2 cache, and the cache line with the parity error is
invalidated by the hardware. All data stored in the L1 data cache is available
in the L2 cache, guaranteeing no data loss.
򐂰 The L2 and L3 caches and the main memory are protected by ECC. In case
of a single bit failure, it is corrected and the system continues to operate
normally. In the case of double bit, errors are detected but not corrected.
򐂰 The L2 cache directory has two parity protected copies of data. If a parity
error is detected, the directory request is recycled while the entry with the
error is refreshed from the good one.
򐂰 The L1 and L2 caches and the L2 and L3 cache directories all have spare bits
in their arrays that can be used to recover from bit failures.
򐂰 Each L3 chip has a spare line to allow for line delete and replacement if the
error becomes hard to recover.
Figure 5-3 graphically represents the redundancy and error recovery
mechanisms on the main memory.
Memory scrubbing for soft single-bit
errors that are corrected in the
background while memory is idle, to help
prevent multiple-bit errors.
Dynamically reassign memory I/O
Failing memory bit steered via bit-steering if error threshold
to spare memory chip
is reached on same bit
Chipkill
X
X
....
Spare
memory
chip
Scatter memory chip bits across four separate ECC words for chipkill recovery
Figure 5-3 Memory error recovery mechanisms
164
IBM ^ pSeries 670 and pSeries 690 System Handbook
The memory implements the following three techniques, which complement the
basic error correction design and significantly improve the robustness.
1. Chipkill™ recovery
Chipkill recovery is accomplished by bit scattering, so that even an entire
DRAM chip failure appears to the ECC logic as correctable single bit errors in
several different ECC words. When a whole memory chip fails, it can be
recovered and replaced by a spare chip.
A technique named bit scattering scatters memory chip bits across four
separate ECC words in order to improve recovery from a memory chip failure.
The failure of any specific memory module only affects a single bit within an
ECC word.
2. Scrubbing
Scrubbing is a hardware memory controller function that determines, during
normal run time, whether or not a correctable permanent failure has occurred
in one or more bit locations. It also reads memory out when it would otherwise
be idle and corrects any single bit errors it detects. When the scrubbing
threshold is reached, the third technique, called redundant bit steering, is
invoked, which steers the data lines away from the faulty chip and onto a
redundant chip. Memory scrubbing takes place while the memory controller is
idle, is completely transparent to software, and uses no CPU cycles.
3. Redundant bit steering
Redundant bit steering is enabled by having the scrubbing hardware detect a
fault, together with steering logic, to provide automatic access to the spare
bits. After redundant bit steering is invoked, the memory region is again free
of permanent bit errors. In case of another bit failure on the same word, the
ECC correction can again be used.
This is named bit steering, and it greatly reduces system outages due to
memory failures.
5.4.3 Multiple data paths
The communication between processors in an MCM is done over four local
buses. Different MCMs are connected through a distributed switch topology,
where the local MCM buses are connected to the buses on the other MCMs, and
data is routed between them. Any processor can access resources connected to
any MCM.
Chapter 5. Reliability, availability, and serviceability
165
The I/O subsystem is based on the Remote I/O link technology. This link uses a
loop interconnect technology to provide redundant paths to I/O drawers. Each I/O
drawer is connected to two RIO ports, and each port can access every
component in the I/O drawer. During normal operations the I/O is balanced
across the two ports. If an RIO link fails, the hardware is designed to
automatically initiate a RIO bus reassignment to route the data through the
alternate path to its intended destination.
Any break in the loop is recoverable using alternate routing through the other link
path and can be reported to the service provider for a deferred repair.
Power to the drawers is controlled from the power controller in the CEC through
the RAS infrastructure (see 2.4.2, “I/O subsystem communication and
monitoring” on page 56).
5.5 Fault recovery
The pSeries 670 and pSeries 690 offer new features to recover from several
types of failure automatically, without requiring a system reboot. The ability to
isolate and deconfigure components while the system is running is of special
importance in a partitioned environment, where a global failure can impact
different applications running on the same system.
Some faults require special handling. We will discuss these in the following
sections.
5.5.1 PCI bus parity error recovery and PCI bus deallocation
PCI bus errors, such as data or address parity errors and timeouts, can occur
during either a DMA operation being controlled by a PCI device, or on a load or
store operation being controlled by the host processor.
During DMA, a data parity error results in the operation being aborted, which
usually results in the device raising an interrupt to the device driver, allowing the
driver to attempt recovery of the operation.
However, all the other error scenarios are difficult to handle as gracefully. On
previous systems, these errors resulted in a bus critical error, followed by a
machine check interrupt and system termination.
166
IBM ^ pSeries 670 and pSeries 690 System Handbook
In the pSeries 670 and pSeries 690, a new I/O drawer hardware, system
firmware, and AIX interaction has been designed to allow transparent recovery of
intermittent PCI bus parity errors, and graceful transition to the I/O device
unavailable state in the case of a permanent parity error in the PCI bus. This
mechanism is known as the PCI Extended Error Handling (EEH).
Standard server PCI implementations can detect errors on the PCI bus, but
cannot limit the scope of the damage, so system termination is often the result.
The I/O drawer (7040-61D) used for the pSeries 670 and pSeries 690 has
enhanced the handling of errors on the PCI bus to limit the scope of the damage
to one PCI slot and loss of use of the PCI adapter in that PCI slot rather than
system termination.
While the basic enablement of PCI EEH is in the circuitry of the I/O drawer,
changes are required in the OS device driver to fully exploit EEH. Many of the
current AIX device drivers are fully enabled for EEH, and most of those that are
not EEH enabled will be enabled in the next AIX release. Also, IBM is working to
have future releases of Linux device drivers EEH enabled.
Use of EEH is particularly important on the pSeries 670 and pSeries 690 running
with multiple partitions. If non-EEH enabled PCI adapters are in use, it is possible
for multiple partitions to be disrupted by PCI bus errors. If a customer’s
application requires the use of non-EEH enabled PCI adapters, careful
placement of those PCI adapters in the I/O drawer can limit PCI bus error
disruptions to a single logical partition.
The ultimate situation is to use only EEH-enabled PCI adapters, to eliminate
system and partition disruptions due to PCI bus errors, and merely suffer the loss
of a single PCI adapter if that adapter causes a PCI bus error. Most adapters
support EEH. It is part of the PCI 2.0 Specification, although its implementation is
not required for PCI compliance.
For those adapters that do not support EEH, special care should be taken when
configuring partitions for maximum availability. See the IBM ^ pSeries
690 Availability Best Practices white paper for I/O adapter configuration
guidelines, which can be found at:
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_avail.html
The principle of how the error is recovered is described as follows:
1. An operation to access a PCI slot causes an error.
2. Upon detection of the error, the access to the bus is denied.
3. The error recovery mechanism probes the PCI bus for the error code.
4. If it is an unrecoverable error, the bus returns 0xFFFFFFFF.
Chapter 5. Reliability, availability, and serviceability
167
5. The device driver performs a call to the firmware requesting a PCI bus reset.
6. After the reset, the device driver retries the operation. If it fails three times, the
slot is deconfigured and a permanent error is logged.
A similar error-handling procedure is performed during system startup when the
PCI bridges are initialized.
5.5.2 Dynamic CPU deallocation
Dynamic CPU deallocation has been available since AIX Version 4.3.3 on
previous RS/6000 and pSeries systems, and is part of the pSeries 670 and
pSeries 690 RAS features. The dynamic CPU deallocation is also supported in a
partitioned environment.
Processors are continuously monitored for errors, such as L1 and L2 cache
errors. When a predefined error threshold is met, an error log with warning
severity and threshold exceeded status is returned to AIX. At the same time, the
service processor marks the CPU for deconfiguration at the next boot. In the
meantime, AIX will attempt to migrate all resources associated with that
processor (threads, interrupts, and so on) to another processor, and then stop
the failing processor.
The typical flow of events for processor deallocation is as follows:
1. The firmware detects that a recoverable error threshold has been reached by
one of the processors.
2. AIX logs the firmware error report in the system error log and, when executing
on a machine supporting processor de-allocation, starts the deallocation
process.
3. AIX notifies non-kernel processes and threads bound to the last logical CPU.
4. AIX waits for all the bound threads to move away from the last logical CPU. If
threads remain bound, AIX eventually times out (after ten minutes) and aborts
the deallocation.
5. Otherwise, AIX invokes the previously registered High Availability Event
Handlers (HAEHs). An HAEH may return an error that will abort the
deallocation.
6. Otherwise, AIX goes on with the deallocation process and ultimately stops the
failing processor.
5.5.3 CPU Guard
It is necessary that periodic diagnostics not run against a processor already
found to have an error by a current error log entry. CPU Guard provides this
168
IBM ^ pSeries 670 and pSeries 690 System Handbook
feature. Otherwise, when an error occurs and the processor cannot be
deallocated, we could end up with multiple periodic diagnostic failures for the
same processor.
Figure 5-4 illustrates how AIX handles the CPU Guard activities.
Periodic Diagnostics finds an
active error log entry in the
OS error log against the processor?
y
n
Periodic Diagnostics runs a
suite of tests against the
processor
Any error
detected?
y
n
Periodic Diagnostics waits a
defined period of time
Periodic Computation
Diagnostics uses AIX facilities
to make an
ibm,os-detected-cpu-fault
RTAS implicating the processor
Figure 5-4 CPU Guard activities handled by AIX
Figure 5-5 illustrates how the firmware also handles those errors.
Chapter 5. Reliability, availability, and serviceability
169
ibm,os-detected-cpu-fault
RTAS Call Seen
Using Service Processor, firmware writes bits
into firmware persistent storage that insures that
firmware will not configure processor on
subsequent reboots.
On next event_scan, RTAS returns an error
indicating that failed CPU should be dynamically
removed from system configuration
When error is logged, OS dynamically
deconfigures CPU using existing mechanisms
for CPU deallocation for predictive CPU failures
Figure 5-5 CPU Guard activities handled by firmware
Enabling CPU Guard
In case of failure at any point of the deallocation, AIX logs the failure and the
reason why the deallocation was aborted. The system administrator can look at
the error log, take corrective action (when possible), and restart the deallocation.
For example, if the deallocation was aborted because at least one application did
not unbind its bound threads, the system administrator could stop the
application(s), restart the deallocation (which should go through this time), and
restart the application. You can turn off the dynamic CPU deallocation by
selecting the following SMIT panels1:
# smit
System Environments
Change / Show Characteristics of Operating System
See the SMIT panel shown in Example 5-1.
1
170
The SMIT shortcut is smitty chgsys.
IBM ^ pSeries 670 and pSeries 690 System Handbook
Example 5-1 smitty chgsys
Change / Show Characteristics of Operating System
Type or select values in entry fields.
Press Enter AFTER making all desired changes.
Maximum number of PROCESSES allowed per user
Maximum number of pages in block I/O BUFFER CACHE
Maximum Kbytes of real memory allowed for MBUFS
Automatically REBOOT system after a crash
Continuously maintain DISK I/O history
HIGH water mark for pending write I/Os per file
LOW water mark for pending write I/Os per file
Amount of usable physical memory in Kbytes
State of system keylock at boot time
Enable full CORE dump
Use pre-430 style CORE dump
CPU Guard
ARG/ENV list size in 4K byte blocks
[Entry Fields]
[128] +#
[1000] +#
[0] +#
false +
false +
[0] +#
[0] +#
1048576
normal
false
false
enable +
[6] +#
+
+
The default value for the CPU Guard flag is enable2 in AIX 5L Version 5.2.
Note: Even if the dynamic deallocation is disabled, AIX still reports the error in
the AIX system error log and you will see the error indicating that AIX was
notified with the problem of a CPU (CPU_FAILURE_PREDICTED).
Sometimes the processor deallocation fails because, for example, an application
did not move its bound threads away from the last logical CPU. Once this
problem has been fixed, by either unbinding (when it is safe to do so) or stopping
the application, the system administrator can restart the processor deallocation
process using the ha_star command, as shown in the following example:
# ha_star -C
ha_star: 1735-008 No previous CPU failure event received.
Use the syntax ha_star -C, where -C is for a CPU predictive failure event. The
processors are represented in the ODM database as procX, where X is the
number of the processor. With the lsattr command you can look at the current
status of a processor, as shown below.
# lsattr -El proc2
state
enable
Processor state False
type
PowerPC_POWER4 Processor type False
frequency 1100264056
Processor Speed False
2
In AIX 5L Version 5.1 or lower, the default value is disable.
Chapter 5. Reliability, availability, and serviceability
171
The state of this processor is enable, so there is no problem encountered so far
with that processor. There are three possible states:
enable
The processor is used by AIX.
disable
The processor has been dynamically deallocated by AIX.
faulty
The processor was declared defective by the firmware at boot time.
5.5.4 Caches and memory deallocation
The L3 cache is ECC protected, so single bit errors are transparently corrected.
When an error occurs, the service processor checks to see if any previous L3
failure occurred and if the failing address matches the previous error. If there is
no match, the service processor will then save the failed address and issue a L3
purge (flush the cache line) to attempt to remove the soft error.
If the address is the same as a previous failure, then this will be treated as a hard
single bit error (SBE), and the service processor has two recovery mechanisms:
Cache line delete
The service processor can use the cache line
delete capability to permanently remove this
cache line from being used. There are two L3
directory cache lines (one per half) that can be
deleted for each L3 module.
L3 cache deconfiguration
If the number of errors exceed two, and both
cache lines are already deleted, the cache
continues to run, but a message calling for
deferred repair is issued. If the machine is
rebooted without such repair, the L3 cache is
placed in bypass mode, and the system comes
up with this cache deconfigured.
The L2 cache can also be deconfigured, as part of the dynamic CPU deallocation
process. If the L2 caches errors inside a POWER4 chip, the CPU deallocation
process is invoked, and the processors and L2 cache are deconfigured from the
system.
Memory cards can be deconfigured at boot time when they fail to pass POST, or
when marked to be deconfigured by the service processor.
5.5.5 Hot-swappable components
Both the pSeries 670 and pSeries 690 provide many parts as hot-swappable
Field Replaceable Units (FRUs). This feature allows you to replace most parts of
the pSeries 670 and pSeries 670 concurrently without the need to power off the
system. There are some parts like the MCMs. L3 cache, memory, etc., that still
172
IBM ^ pSeries 670 and pSeries 690 System Handbook
require a scheduled maintenance window to perform the replacement. Table 5-1
provides you with an overview of which components are hot-pluggable.
Table 5-1 Hot-swappable FRUs
Processor subsystem FRUs
Hot-swappable concurrent maintenance
Blowers
Yes
DCA
Yes
Bulk power enclosure
Yes
Bulk power controller (BPC)
Yes
Bulk power regulator (BPR)
Yes
Bulk power distributor (BPD)
Yes
Bulk power fan (BPF)
Yes
UEPO switch panel
Yes
Internal battery feature (IBF)
Yes
Capacitor book
No
Processor subsystem chassis
No
MCM
No
Memory cards
No
L3 cache
No
Clock card
No
I/O books
No
I/O subsystem FRUs
Hot-swappable concurrent maintenance
I/O backplane and riser card
No
DASD backplane
No
Disk drives
Yes
DCA (power supplies)
Yes
I/O fan assemblies
Yes
PCI adapters
Yes
Media subsystem FRUs
Operator panel
Hot-swappable concurrent maintenance
No
Chapter 5. Reliability, availability, and serviceability
173
Processor subsystem FRUs
Hot-swappable concurrent maintenance
Diskette drive
Yes
CD-ROM/DVD-RAM/DVD-ROM
Yes
Optional media SCSI devices
Yes
5.5.6 Hot-swappable boot disks
The I/O drawer (7040-61D) of the pSeries 670 and pSeries 690 provides up to 16
hot-swappable bays for internal disks, organized in four cages, each one
connected to a separate Ultra3 SCSI controller.
Disk mirroring is strongly suggested for the operating system and paging in
order, to eliminate system outages because of operating system disk failures. If a
mirrored disk fails, AIX automatically reroutes the I/O requests to the other disk,
without service interruption. You can replace the failed drive with the system
online, and mirror again, to restore disk redundancy.
5.5.7 Hot-Plug PCI adapters
All PCI slots on the pSeries 670 and pSeries 690 are PCI 2.2-compliant and are
hot-plug enabled, which allows most PCI adapters to be removed, added, or
replaced without powering down the system. This function enhances system
availability and serviceability.
The function of Hot-Pluggable PCI adapters is to provide concurrent adding or
removal of PCI adapters when the system is running. In the I/O drawer, the
installed adapters are protected by plastic separators called blind swap cassettes.
These are used to prevent grounding and damage when adding or removing
adapters.
Figure 5-6 shows a drawing of the blind swap cassette.
174
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 5-6 Blind swap cassette
The Hot-Plug LEDs outside the I/O drawer indicate whether an adapter can be
plugged into or removed from the system. The Hot-Plug PCI adapters are
secured with retainer clips on top of the slots; therefore, you do not need a
screwdriver to add or remove a card, and there is no screw that can be dropped
inside the drawer. Just in case of exchanging an PCI adapter from the blind swap
cassette, a screwdriver is needed to remove/replace it from the cassette. To fit
several adapters that are different sizes, there are several different blind swap
cassettes available.
Figure 5-7 shows a schematic of an I/O drawer with the status LEDs for the RIO
ports and PCI slots.
Chapter 5. Reliability, availability, and serviceability
175
Figure 5-7 Hot-Plug PCI adapters, blind swap cassette, and status LEDs
The function of Hot-Plug is not only provided by the PCI slot, but also by the
function of the adapter. Most adapters are Hot-Pluggable, but some are not. Be
aware that some adapters must not be removed when the system is running,
such as the adapter with the operating system disks connected to it or the
adapter that provides the system console.
For further information, refer to AIX 5L Version 5.2 System Management Guide:
Operating System and Devices (located on the documentation CD-ROM that
ships with the AIX operating system) and from PCI Adapter Placement
References, SA38-0538.
5.5.8 Light Path Diagnostics
The pSeries 670 and pSeries 690 introduce a new concept named Light Path
Diagnostics, which provides a visual identification of a failed component in order
to facilitate the maintenance. This functionality is available for detecting problems
in I/O drawers, PCI planers, fans, blowers, and disks.
When a component presents a problem and it is detected by the surveillance
mechanisms, the physical location occupied by the failed component flashes a
LED, allowing easy identification of the component requiring maintenance.
176
IBM ^ pSeries 670 and pSeries 690 System Handbook
Disk drives, fans, blowers, and PCI adapters have specific LEDs to indicate the
operational state and faults when they occur. Figure 5-8 shows the LED
placement on I/O drawers from the front view.
Blower1A
Blower 2A
Blower1BA
Blower 2B
Power Good Out
Drawer Fault/Identify
DCA power Fault
Power Good In
Four pack
Fan
Fault
LED
Light Path Diagnostics(tm)
with LEDs for:
Fans, blowers,
power, disk removal,
and faults.
Backplane
Fault
LED
Green = Activity
Amber = Identify/Remove/Fault
Figure 5-8 Status LEDs for components in I/O drawers
The Hardware Service Functions available in the Service Focal Point application
on the HMC provides you with the method to identify the LEDs for the frame,
cards, and power supplies (see “How to use Hardware Service Function” on
page 199).
5.6 Serviceability features
The serviceability features on the pSeries 670 and pSeries 690 are centralized
on the HMC and provided under Software Maintenance (see Figure 5-9 on
page 178) and Service Applications in the Navigation Area (see Figure 5-10 on
page 179).
Chapter 5. Reliability, availability, and serviceability
177
Figure 5-9 Software maintenance
Note: “Microcode Update” will be shown only when HMC software release
Version 1.3.2 or later is installed.
Software Maintenance contains the following applications:
򐂰 Frames3
򐂰 HMC
򐂰 Microcode Updates
In 5.6.2, “Upgrading HMC” on page 181 we describe the steps involved to
upgrade the HMC software via the HMC application, and in 5.6.3, “Microcode
Updates function” on page 184 we describe the steps involved to update the
microcodes via the Microcode Updates application.
3
178
Frame functions will only be available in the future.
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 5-10 Service Applications
Service Applications contain the following applications:
򐂰
򐂰
򐂰
򐂰
Problem Determination4
Inventory Scout Services
Service Agent
Service Focal Point
Figure 5-11 illustrates the relationship between these components.
4
The problem determination function is only available for the product support engineers.
Chapter 5. Reliability, availability, and serviceability
179
CEC
Service Processor
Cycles continuously
looking for faults
AIX Error Log
Input/Output
No
(R ti fi e
S/
23 s
2C
)
AIX device drivers
provide error
information to the
AIX error log
Partition
or
s
IBM.ServiceRM in AIX
Notifies error
information to
the HMC
No
tif
Mo
nit
ie
fra
s
m
ew
or
k)
Central
repository
for all
system
errors
HMC
Service
Focal Point
On-board in the CEC.
Runs asynchronously.
Does not degrade CEC
performance.
(R
M
C
Processors
Memory
Power
Thermal
Blowers/Fans
IBM Support Center
Initiates maintenance
action with the parts
required to fix the real
problem
Figure 5-11 Error reporting and consolidation
All the applications’ usage is well explained in IBM Hardware Management
Console for pSeries Installation and Operations Guide, SA38-0590, therefore, we
focused on the following points in this section:
򐂰 What components comprise these applications
򐂰 Problem determination hints
5.6.1 Back up of HMC
There are two types of backup that you need to be aware of: Backup Critical
Console Data and Save Upgrade Data.
For Backup Critical Console Data, its function is to provide a means for
customers to save the data on the HMC that has changed since the
creation/installation of the code on the HMC to the DVD-RAM. All files including
those belonging to packages installed after the date the HMC Recovery image
was created will be backed up. The typical types of files that are backed up with
this function are backup copies of LPAR/Profiles, system information such as
network configurations, user configuration file and individual user’s home
directory. The backup function should be performed each time configuration
changes are made to the system or the HMC.
180
IBM ^ pSeries 670 and pSeries 690 System Handbook
Note: If the HMC software is upgraded from one release to another, do not
use the Backup Critical Console Data to restore the configurations. This back
up will restore the configurations including the previous HMC release.
For Save Upgrade Data, this function should only be used when customer wants
to upgrade the current code on their HMC from one release/version to a higher
release/version with a new HMC Recovery CD. This is not required if the code
from the new release is downloaded/installed from the Support Web site or from
an update CD ordered from IBM. Data saved by this function resides only in a
special partition on the HMC disk. After the installation of the new HMC software
from the HMC Recovery CD, on the first reboot, a service will examine the
partition if there is any data, and then restore the data. For more information,
refer to IBM Hardware Management Console for pSeries Installation and
Operations Guide, SA38-0590.
5.6.2 Upgrading HMC
.
Note: If you have an HMC with a release earlier than Release 3, please read
the chapter about “Installing and Configuring the HMC” in the IBM Hardware
Management Console for pSeries Installation and Operations Guide,
SA38-0590.
Here are the steps for updating HMC software from V1.3.0 to V1.3.2:
1. Before the upgrade, verify the current HMC software version by selecting
Software Maintenance -> HMC as shown in Figure 5-12.
Chapter 5. Reliability, availability, and serviceability
181
Figure 5-12 Software Maintenance: HMC
2. Select Format Removable Media (see circle section in Figure 5-12) -> DVD
RAM (see Figure 5-13) to format the DVD RAM which will be used for back
up.
Figure 5-13 Format Media
3. Select Backup Critical Console Data to back up the data onto the DVD RAM
(see Figure 5-14).
182
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 5-14 Backup Critical Data
4. Select Install Corrective Service to upgrade the HMC software. The user
needs to select where the corrective service files are located, which can be
either on a CD or a remote site. For this example, the files are located at a
remote site (see Figure 5-15).
Figure 5-15 Install Corrective Service
5. After the upgrade is completed, if the managed systems are connected
through 8-port or 128-port asynchronous adapters, you would need to
reconfigure these serial adapters. The details can be found in the chapter
about “Using One HMC to Connect to More than One Managed System” in
Chapter 5. Reliability, availability, and serviceability
183
the IBM Hardware Management Console for pSeries Installation and
Operations Guide, SA38-0590.
6. Reboot the HMC and verify that the HMC software is upgraded correctly by
selecting Software Maintenance -> HMC (see Figure 5-16).
Figure 5-16 Software Maintenance: HMC (2)
5.6.3 Microcode Updates function
There are several aspects to the overall microcode management strategy which
include the survey, distribution and installation of microcodes. In the past, it
would required an IBM Customer Engineer (CE) to go to the customer’s site to
survey and update the microcodes. With this Microcode Updates function
available in the HMC and AIX, the customer can now manage their microcodes
without the need for a CE.
Figure 5-17 on page 185 shows how the mechanism of the microcode update
works from a stand-alone AIX server and from an HMC. A single application
called Inventory Scout is used to manage the microcodes from both AIX and
HMC. The application is able to survey and report the existing and latest
microcode levels, and provide suggested actions for each device. It also provides
184
IBM ^ pSeries 670 and pSeries 690 System Handbook
the capability for installation of microcodes and warns the user before a reboot or
installation of back level codes. With this, the customer is able to keep current on
the microcode levels at their convenience.
Note: In order to have the Microcode Updates function:
򐂰 For a stand-alone AIX server, the operating system must be either be
5100-04 Recommended Maintenance Level, 5200-01 Recommended
Maintenance Level, or later.
򐂰 For HMC, the HMC software must be V1.3.2 or later.
HMC
AIX
WebSM GUI
Standalone
WebSM GUI
WebSM Server
WebSM Server
Switches
DCA
BPA
Inventory
Scout
Device
Firmware
System
Firmware
Inventory
Scout
RMC
Partition
Service
Partition
Partition
Inv Scout
Inv Scout
Inv Scout
Device
Firmware
Device
Firmware
Device
Firmware
System Firmware
Figure 5-17 Mechanism of the Microcode Updates
Here are the steps to survey and install the latest microcode levels from the
HMC:
1. Select Software Maintenance -> MicroCode Updates, as shown in
Figure 5-18.
Chapter 5. Reliability, availability, and serviceability
185
Figure 5-18 Microcode Updates
2. Select Change Location if you want to change from the default Web site
Service location, as shown in Figure 5-19.
Figure 5-19 Download and Apply Microcode
186
IBM ^ pSeries 670 and pSeries 690 System Handbook
3. Select the location where the latest microcode levels are stored and select
OK as shown in Figure 5-20.
Figure 5-20 Select Repository Location
4. After entering the location, highlight the systems that you want to survey and
select Survey. When the survey is completed, a summary of all the devices,
their current and latest microcode levels, effects of the updates and
suggested actions will be displayed as shown in Figure 5-21.
Note: For any selected devices, based on the “Effect” column:
򐂰 “Take offline” - The user must take devices offline prior the update of the
microcode, otherwise the update will fail.
򐂰 “Reboot” - The user must confirm that he understands that the system will
reboot as a result of the microcode updates. Applications should be
stopped, all users should be notified and the non service partitions should
be shut down.
Chapter 5. Reliability, availability, and serviceability
187
Figure 5-21 Microcode Survey Results
5. It is also possible to view the details for each device. Highlight the device to be
viewed and select View Information. The details of the device will then be
shown as in Figure 5-22.
Figure 5-22 Microcode Installation - Device Installation
188
IBM ^ pSeries 670 and pSeries 690 System Handbook
Note: It is only possible to view one device at a time.
6. The user can either update all or some of the devices by selecting on the
“Install” check box on the device line(s) and click Apply as shown in
Figure 5-21.
A warning message appears as in Figure 5-23 to warn you about the effects
of the updates for system firmware updates. The user can either proceed with
the update by selecting OK or abort the update by selecting Cancel.
Figure 5-23 Warning message
7. You can monitor the Microcode Updates process from a menu as shown in
Figure 5-24.
Chapter 5. Reliability, availability, and serviceability
189
Figure 5-24 Updating microcodes
8. When the microcode update is completed, a window appears as shown in
Figure 5-25. The system will automatically reboot if required.
190
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 5-25 Microcode Updates completed
5.6.4 Inventory Scout Services
Inventory Scout Services are pSeries tools that perform the following two
functions:
򐂰 Microcode Discovery Service
This generates a real-time comparison report showing subsystems that may
need to be upgraded. This microcode survey function is similar to the
microcode survey feature discussed in 5.6.3, “Microcode Updates function”
on page 184, but without the updates capability. For further information about
Microcode Discovery Service, visit the following URL:
http://techsupport.services.ibm.com/server/aix.invscoutMDS
򐂰 VPD Capture Service
This transmits your server’s vital product data (VPD) information to IBM. For
further information about VPD Capture Service, visit the following URL:
http://techsupport.services.ibm.com/server/aix.invscoutVPD
You can perform these operations by selecting the following two tasks, as shown
in Figure 5-26:
򐂰 Conduct Microcode Survey
Chapter 5. Reliability, availability, and serviceability
191
򐂰 Collect VPD Information
Figure 5-26 Inventory Scout Services
By selecting these two tasks on the HMC, you can check the managed system
for needed microcode updates and collect vital product data (VPD) from
partitions or the Full System Partition. If Service Agent is configured on the HMC,
these files are automatically sent to IBM. You can optionally save files on
DOS-formatted diskettes on the HMC diskette drive.
AIX 5L Version 5.2 includes the following Inventory Scout file sets5:
# lslpp -L invscout.*
Fileset
Level State Type Description (Uninstaller)
---------------------------------------------------------------------------invscout.ldb
1.5.0.0
C
F
Inventory Scout Logic
Database
invscout.msg.en_US.rte
1.2.0.0
C
F
Inventory Scout Messages U.S. English
invscout.rte
1.5.0.0
C
F
Inventory Scout Runtime
You can invoke the invscout command on AIX using the command line interface
to Inventory Scout, as shown below.
5
192
AIX 5L Version 5.1 includes a different level of file sets.
IBM ^ pSeries 670 and pSeries 690 System Handbook
# /usr/sbin/invscout
****** Command ---- V1.5.0.0
****** Logic Database V1.5.0.0
Initializing ...
Identifying the system ...
Getting system firmware level(s) ...
Scanning for device firmware level(s) ...
64 devices detected; each dot (.)
represents 10 devices processed:
......
Writing Microcode Survey upload file ...
Microcode Survey complete
The output files can be found at:
Upload file: /var/adm/invscout/lpar04.mup
Report file: /var/adm/invscout/invs.mrp
To transfer the invscout 'Upload file' for microcode
comparison, see your service provider's web page.
In fact, this is equivalent to the Conduct Microcode Survey task on the HMC, and
the command examines the microcode levels of all hardware components visible
in the partition, then accumulates the results in the microcode survey upload file.
The brief usage of invscout is shown below.
# invscout -h
Usage:
invscout [-v] [-c] [-r] [-m] type-model [-s] serial [-g] [-q] [-k] [-h]
-v
-c
-r
-m
-s
-g
-- Change the survey type from
'Microcode' to 'VPD'
-- Change the main action from 'perform a
new survey' to 'concatenate existing
survey upload files'
-- For a new Microcode Survey, sends a
copy of the formatted text report
file to the screen from which the
command was invoked
-- Allows input of the platform machine
type and model for VPD surveys
-- Allows input of the platform serial
number for VPD surveys
-- Displays the versions of this command
Chapter 5. Reliability, availability, and serviceability
193
-q
-k
-h
and the logic database currently in
use
-- Suppresses most run-time messages
-- Keeps temporary files when the
command is complete
-- Generates this usage statement
On the pSeries 670 and pSeries 690 managed by an HMC, the accumulated file
can be sent to IBM using the following steps:
1. The file is transferred to the HMC from the partition.
2. The file is sent to IBM from the HMC.
Transferring files to the HMC from partitions
If the following condition is met, files will be automatically transferred to the HMC
from the partition after the initial automatic configuration is made on the HMC.
򐂰 The firmware level of pSeries 670 or pSeries 690 is 10/2002 system
microcode update or later.
򐂰 The software level of HMC is Release 3 or higher.
򐂰 The partition is installed with either of the following:
– AIX 5L Version 5.1 with 5100-03 Recommended Maintenance Level
– AIX 5L Version 5.2
Otherwise, Inventory Scout uses its own authentication method between the
partition and the HMC in order to talk to the Inventory Scout daemon (invscoutd)
on AIX. Therefore, it requires the following additional setup on AIX in a partition:
򐂰 A user, invscout, must be defined on the partition.
򐂰 A password (for example, invscout) must be set for the invscout user.
To configure Inventory Scout on partitions, do the following:
1. In the Navigation area, click the Service Applications icon.
2. In the Contents area, double-click the Inventory Scout Services icon.
3. In the Contents area, click Inventory Scout Profile Configuration, shown in
Figure 5-26. The Inventory Scout Configuration Assistant window will open.
4. Select the managed system you want to configure and click Next.
194
IBM ^ pSeries 670 and pSeries 690 System Handbook
5. A list of partitions, along with each partition’s configuration status, will be
displayed, as shown in Figure 5-27.
Figure 5-27 Inventory Scout Configuration Assistant
If one of active partitions shows automatically configured, then the partition is
already configured. If one of the active partitions shows not configured, instead
of automatically configured, you must manually configure the Inventory Scout
service for that partition. To configure, select the partition (shown in Figure 5-27)
and click Next, then enter the following information (the host name or IP address
of the partition will be filled automatically):
򐂰 The password of the invscout user on that partition
򐂰 The listening port of invscoutd (default value is 808)
5.6.5 Service Agent
Electronic Service Agent (also known as Service Agent) is an application
program that runs on either AIX or Linux6 to monitor the system for hardware
errors. On pSeries systems managed by the HMC, the primary path for system
hardware errors detection and analysis consists of the diagnostics function
provided by AIX, the service processor, and the Service Focal Point (see
“Service Agent” on page 195). Service Agent provides the transport facility to
IBM.
6
Service Agent supports the Linux operating system on HMC only.
Chapter 5. Reliability, availability, and serviceability
195
Service Agent can execute several tasks, including:
򐂰
򐂰
򐂰
򐂰
򐂰
Automatic problem analysis
Problem-definable threshold levels for error reporting
Automatic problem reporting
Automatic customer notification
Visualize hardware error logs
Although there are several scenarios available for the Service Agent network
configuration, we only explain the configuration, which is used on the HMC with
managed systems, shown in Figure 5-28.
pSeries 670 or pSeries 690
in a partitioned environment
partition
(AIX)
HMC (Linux)
Service Focal Point
GUI
partition
(AIX)
IBM.ServiceRM
Service Agent
gateway process
Dial up
connection
Modem
IBM
Support
Center
IBM.ServiceRM resource manager running on AIX in a partition
Figure 5-28 Service Agent on the HMC
In this configuration, the Service Agent gateway process running on the HMC
places service calls to the IBM support center via a dial-up connection with the
attached modem, if necessary.
Note: No human intervention is required for this process.
By utilizing Service Agent, the pSeries 670 and pSeries 690 can reduce the
amount of downtime experienced in the event of a system component failure by
giving the service provider the ability to view the error report entry and, if needed,
order any necessary replacement parts prior to arriving on site. The opportunity
for human misinterpretation or miscommunication in problem determination is
therefore mitigated.
Figure 5-29 shows the initial configuration panel of Service Agent.
196
IBM ^ pSeries 670 and pSeries 690 System Handbook
Figure 5-29 Service Agent on the HMC
For more information on the Service Agent, please refer to the following
publications:
򐂰 Electronic Service Agent for pSeries and RS/6000 User’s Guide, available at:
ftp://ftp.software.ibm.com/aix/service_agent_code/AIX/svcUG.pdf
򐂰 Electronic Service Agent for pSeries Hardware Management Console User’s
Guide, available at:
ftp://ftp.software.ibm.com/aix/service_agent_code/HMC/HMCSAUG.pdf
򐂰 IBM Hardware Management Console for pSeries Installation and Operations
Guide, SA38-0590
Chapter 5. Reliability, availability, and serviceability
197
5.6.6 Service Focal Point
Traditional service strategies become more complicated in a partitioned
environment. Each partition runs on its own, unaware that other partitions exist
on the same system. If one partition reports an error for a shared resource, such
as a managed system power supply, other active partitions report the same error.
To enable service representatives to avoid long lists of repetitive call-home
information, the HMC provides the Service Focal Point application. Service Focal
Point recognizes that these errors repeat, and filters them into one serviceable
event for the service representative to review.
The Service Focal Point is a system infrastructure on the HMC that manages
serviceable event information for the system building blocks. It includes resource
managers that monitor and record information about different objects in the
system. It is designed to filter and correlate events from the resource managers
and initiate a call to the service provider when appropriate. It also provides a user
interface that allows a user to view the events and perform problem
determination.
Note: Service Focal Point only collects hardware errors, such as
PERMANENT errors from AIX (marked as P) and NON BOOT errors from the
service processor.
Upon hardware failure events, the corresponding error entry is notified from the
partition to the HMC, as shown in Figure 5-11 on page 180. The IBM.ServiceRM
subsystem is in charge of this notification. The AIX diagnostic function creates a
serviceable event through IBM.ServiceRM when a hardware problem is
determined and events will be notified to the HMC using the Resource
Monitoring and Control (RMC) framework. The IBM.ServiceRM is running as
the IBM.ServiceRMd daemon and packaged in the
devices.chrp.base.ServiceRM file set in AIX, as shown below.
# lssrc -g rsct_rm | head -1; lssrc -g rsct_rm | grep ServiceRM
Subsystem
Group
PID
Status
IBM.ServiceRM
rsct_rm
307354
active
# ps -ef | head -1; ps -ef | grep ServiceRM | grep -v grep
UID
PID PPID
C
STIME
TTY TIME CMD
root 307354 122982 0 Sep 11
- 0:31 /usr/sbin/rsct/bin/IBM.ServiceRMd
# lslpp -w /usr/sbin/rsct/bin/IBM.ServiceRMd
File
Fileset
Type
---------------------------------------------------------------------------/usr/sbin/rsct/bin/IBM.ServiceRMd
devices.chrp.base.ServiceRM
File
# lslpp -L devices.chrp.base.ServiceRM
Fileset
Level State Type Description (Uninstaller)
----------------------------------------------------------------------------
198
IBM ^ pSeries 670 and pSeries 690 System Handbook
devices.chrp.base.ServiceRM
1.2.0.0
C
F
RSCT Service Resource
Manager
From the Service Focal Point interface, you can execute maintenance
procedures such as examining the error log history, checking for components
requiring replacement, and performing a Field Replaceable Unit (FRU)
replacement. If Service Agent is configured on the HMC, the serviceable events
are automatically sent to IBM (call-home support) for automatic generation of a
maintenance request.
How to use Hardware Service Function
This function allows you either to just identify a frame when you have several
frames connected to your HMC, or to turn off the rack indicator light. You are also
able to get a Field Replaceable Unit (FRU) list when the rack indicator light is lit
and check which component has problems. When a component is shown here
with the LED state ON, it is much easier to identify the failing component.
To use this function, do the following:
1. Click the + mark left of the “Service Applications” in the Navigation Area.
2. Select Service Focal Point in the Navigation Area. You will see the Service
Focal Point task panel, as shown in Figure 5-30.
3. Select Hardware Service Functions (Figure 5-30).
Figure 5-30 Service Focal Point: Hardware Service Functions
Chapter 5. Reliability, availability, and serviceability
199
4. You will see the Hardware Service Management: Overview window, as shown
in Figure 5-31. Select the managed system for which you want to check the
LED state, then select the List FRUs button.
Figure 5-31 Hardware Service Functions overview
5. You will see the FRU LED Management window (Figure 5-32).
Figure 5-32 FRU LED Management
200
IBM ^ pSeries 670 and pSeries 690 System Handbook
If any of the LEDs are ON, it would mean that the system has a problem with the
indicated component. If the Service Agent is configured to notify IBM of the
errors to IBM, then IBM customer service representatives will be informed of the
problem.7
For more information, please refer to IBM Hardware Management Console for
pSeries Installation and Operations Guide, SA38-0590.
5.6.7 Problem determination hints of Service Functions
To prevent problems with the configuration of Service Functions on the HMC,
such as Inventory Scout, Service Agent, and Service Focal Point, the Ethernet
network between the HMC and partitions, there should be a mandatory
administrative network used by only these applications.
With careful network planning, you should not have any problems using these
applications; however, if an AIX administrator mistakenly changes the TCP/IP
configuration on a partition without notifying the HMC administrator, it might
result in severe communication problems.
Authentication mechanism
The Service Focal Point and DLPAR functions rely on the RMC framework
between the HMC and partitions. The RMC framework performs not only session
management, but also authentication between network peers.
The ctcas subsystem, also known as the cluster authentication daemon, is in
charge of this authentication mechanism. It is running as the ctcasd daemon and
is packaged in the rsct.core.sec file set in AIX, as shown below.
# lssrc -g rsct
Subsystem
Group
PID
Status
ctrmc
rsct
299204
active
ctcas
rsct
188658
active
# ps -ef | head -1; ps -ef | grep ctcas | grep -v grep
UID
PID PPID
C
STIME
TTY TIME CMD
root 188658 139350
0
Sep 11
- 0:03 /usr/sbin/rsct/bin/ctcasd
# lslpp -w /usr/sbin/rsct/bin/ctcasd
File
Fileset
Type
---------------------------------------------------------------------------/usr/sbin/rsct/bin/ctcasd
rsct.core.sec
File
The configuration process of authentication between the HMC and partitions can
be briefly summarized as shown in Table 5-2.
7
In order to dispatch IBM customer service representatives, you need the maintenance agreement
(MA) for this system.
Chapter 5. Reliability, availability, and serviceability
201
Table 5-2 Authentication process
Sequence
On the HMC
1
The DMSRM resource manager places
the secrete key and the HMC host
name in the NVRAM of the managed
system. Every reboot of the HMC, it
places a new secret key.
On an AIX partition
2
The IBM.CSMAgentRM resource manager
reads the secret key and the HMC host name
from NVRAM using an RTAS call. The NVRAM is
checked every five minutes to detect new
HMC(s) and/or key changes. An existing HMC
with a changed key will cause the registration
process (the next 2 steps) to be performed
again.
3
Once the HMC and LPAR have authenticated
each other using the secret key and have
exchanged some information about each other,
(for example, public keys), IBM.CSMAgentRM
grants the HMC permission to access the
necessary resource classes on the partition.
Without proper permission on AIX, the HMC will
be able to establish a session with the partition,
but will not be able to query for the operating
system information, such as DLPAR capabilities,
or execute DLPAR operation commands
afterward.
4
The last part of the registration process is the
creation of an IBM.ManagedNode resource with
a Hostname attribute set to the partition's host
name on the HMC. Then, an
IBM.ManagementServer resource will be
created with a Hostname attribute set to the
HMC's host name on the partition.
5
After the ManagedNode resource is
created and authenticated, the
ServiceRM and LparCmdRM resource
managers establish a session with the
partition for DLPAR operation and
receive serviceable events.
202
IBM ^ pSeries 670 and pSeries 690 System Handbook
Note: The current implementation of authentication mechanism used in the
RMC framework is called UNIX hostname authentication. The RMC, and
therefore the HMC, may implement new authentication mechanisms in
accordance with the future development plan of RMC.
Trouble-free network planning rules
To avoid unnecessary configuration errors in Service Function applications, you
must understand the following rules:
򐂰 All the combinations of a host name and an IP address must be unique.
򐂰 All the network interfaces on the HMC and partitions must be assigned
different host names and, therefore, different IP addresses.
򐂰 The assigned IP address must be consistently resolved regardless the
location (on the HMC or partitions). If some name services, such as NIS,
DNS, and LDAP, are used, they must be reliable and return the consistent
results.
򐂰 The network interface on the HMC, which is resolved to the node name (the
string returned from the hostname command), must be reachable from all the
partitions.
The following examples show inappropriate network configurations.
򐂰 Duplicate IP addresses
Two partitions have different host names, but the same IP address on their
network interface.
򐂰 Unresolvable host name
A partition does not have the valid DNS configuration, while the HMC uses
DNS for the name resolution. The partition cannot resolve the HMC’s host
name to an IP address (unresolvable).
򐂰 Inconsistent name resolution
The HMC is assigned the fully qualified domain name (FQDN)
itsohmc.itsc.austin.ibm.com for both node name and the host name for eth0
interface. An AIX partition uses DNS for the name resolution, but there are the
following files on the partition:
# cat /etc/netsvc.conf
hosts=local,bind
# grep itsohmc /etc/hosts
9.3.4.30
itsohmc
itsohmc.itsc.austin.ibm.com
Chapter 5. Reliability, availability, and serviceability
203
Therefore, the same IP address 9.3.4.30 is resolved as:
On the HMC
itsohmc.itsc.austin.ibm.com
On the partition
itsohmc
򐂰 Unreachable network interface
The HMC has two network interfaces, eth0 and eth1. Although the FQDN
itsohmc.itsc.austin.ibm.com is assigned for both node name and host name
for the eth0 interface, all partitions can reach to eth1 interface only.
We strongly suggest that you do the following before doing any recovery
activities:
1. Issue the hostname command on the HMC and all partitions. To issue the
hostname command on the HMC, you can use OpenSSH, as shown in the
following example:
$ whence ssh
/usr/bin/ssh
$ ssh -l hscroot itsohmc.itsc.austin.ibm.com hostname
[email protected]'s password: XXXXXX
itsohmc.itsc.austin.ibm.com
For further information about how to use OpenSSH on AIX, please refer to
Managing AIX Server Farms, SG24-6606.
2. Issue the host command against all of the network interfaces on the HMC and
all of the partitions.
a. Confirm how many interfaces are available.
$ ssh -l hscroot itsohmc.itsc.austin.ibm.com\
“/sbin/ifconfig -l | grep Link”
[email protected]'s password: XXXXXX
eth0
Link encap:Ethernet HWaddr 00:02:55:13:85:2E
lo
Link encap:Local Loopback
b. Confirm the IP address of eth0.
$ ssh -l hscroot itsohmc.itsc.austin.ibm.com /sbin/ifconfig eth0
[email protected]'s password: XXXXXX
eth0
Link encap:Ethernet HWaddr 00:02:55:13:85:2E
inet addr:9.3.4.30 Bcast:9.3.5.255 Mask:255.255.254.0
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:1256676 errors:0 dropped:0 overruns:0 frame:7
TX packets:1381966 errors:0 dropped:0 overruns:0 carrier:13
collisions:404844 txqueuelen:100
RX bytes:132305448 (126.1 Mb) TX bytes:1048151698 (999.5 Mb)
Interrupt:10 Base address:0x5000
204
IBM ^ pSeries 670 and pSeries 690 System Handbook
c. Confirm both the reverse and regular name resolutions.
# ssh -l hscroot itsohmc.itsc.austin.ibm.com host 9.3.4.30
[email protected]'s password: XXXXXX
30.4.3.9.in-addr.arpa. domain name pointer itsohmc.itsc.austin.ibm.com.
# ssh -l hscroot itsohmc.austin.ibm.com host itsohmc.itsc.austin.ibm.com
[email protected]'s password:
itsohmc.itsc.austin.ibm.com. has address 9.3.4.30
Changing the host name manually
To change the host name on your partition manually without having any problems
with Service Functions applications, do the steps below on the partition.
Note: These steps are not applicable on the partition installed with AIX 5L
Version 5.1 with 5100-03 Recommended Maintenance Level or AIX 5L
Version 5.2.
The steps are:
1. If you are running AIX 5L Version 5.1 in the partition without applying 5100-03
Recommended Maintenance Level, then do the following:
# /usr/sbin/rsct/bin/runact -c IBM.ManagementServer SetRTASPollingInterval\
Seconds=0
2. Run the following command:
# /usr/sbin/rsct/bin/lsrsrc IBM.ManagementServer Hostname
You will receive output similar to the following example (there may be multiple
entries if the partitions are managed by multiple HMCs):
Resource Persistent Attributes for: IBM.ManagementServer
resource 1:
Hostname = "itsohmc.itsc.austin.ibm.com"
3. The rmrsrc command is now used to delete the resources using the host
name shown before. Repeat this step for each entry (using its Hostname) and
verify none remain with the command shown in step 2.
# /usr/sbin/rsct/bin/rmrsrc -s'Hostname = "itsohmc.itsc.austin.ibm.com"'\
IBM.ManagemementServer
4. Stop the RMC subsystem using the rmcctrl command:
# /usr/sbin/rsct/bin/rmcctrl -z
5. Change the partition host name.
6. Restart the RMC subsystem:
# /usr/sbin/rsct/bin/rmcctrl -A
0513-071 The ctrmc Subsystem has been added.
Chapter 5. Reliability, availability, and serviceability
205
0513-071 The ctcas Subsystem has been added.
0513-059 The ctrmc Subsystem has been started. Subsystem PID is 19628.
You can verify with the command in step 2 whether the host name of the HMC is
filed again in the resource configuration database.
For more information on the Inventory Scout Services, please refer to IBM
Hardware Management Console for pSeries Installation and Operations Guide,
SA38-0590.
For further information about the RMC framework and its resource managers,
please refer to the following publications:
򐂰 A Practical Guide for Resource Monitoring and Control (RMC), SG24-6606
򐂰 IBM Reliable Scalable Cluster Technology for AIX 5L RSCT Guide and
Reference, SA22-7889
򐂰 IBM Reliable Scalable Cluster Technology for AIX 5L Messages, GA22-7891
򐂰 IBM Reliable Scalable Cluster Technology for AIX 5L Technical Reference,
SA22-7890
5.7 AIX RAS features
The RAS features described in the previous sections are all based on the
pSeries 670 and pSeries 690 hardware. There are some additional RAS features
that work in conjunction with the AIX operating system, and some that are
entirely dependent on AIX. These features are covered in the following sections.
5.7.1 Unrecoverable error analysis
In prior RS/6000 and pSeries systems, data corruption due to parity or cyclic
redundancy check (CRC) errors on buses as well as ECC uncorrectable errors in
all levels of cache and memory usually resulted in a machine crash. The pSeries
670 and pSeries 690 adds a new functionality so that these errors are captured
and sent to the partition affected by the failed resources, and provide the
operating system with the error code for analysis.
AIX analyzes the results of the firmware recovery attempt and determines which
process is corrupted by the uncorrectable hardware error. If the process is a user
process, it will be terminated. If the process is within the AIX kernel, then the
operating system will be terminated. Again, terminating AIX in a partition will not
impact any of the other partitions.
206
IBM ^ pSeries 670 and pSeries 690 System Handbook
5.7.2 System hang detection
AIX 5L Version 5.1 and higher offers a feature called system hang detection that
provides a mechanism to detect system hangs and initiates a pre-configured
action. It relies on a new daemon named shdaemon, and a corresponding
configuration command named shconf.
In a case where applications adjust their processes or thread priorities using
system calls, there is a potential problem that their priorities will become so high
that regular system shells are not scheduled. In this situation, it is difficult to
distinguish a system that really hangs (it is not doing any meaningful work
anymore) from a system that is so busy that none of the lower priority tasks, such
as user shells, have a chance to run.
The new system hang detection feature uses a shdaemon entry in the /etc/inittab
file with an action field that specifies what should be done when certain
conditions are met. The actions can be, for example, generating an entry in the
error log, displaying a warning message, running a predefined command,
opening a high-priority login prompt on tty0, or restarting the system.
For more information on how to configure this feature, please refer to Managing
AIX Server Farms, SG24-6606.
5.7.3 AIX disk mirroring and LVM sparing
Mirroring the operating system boot disks (called rootvg mirroring) is a feature
available on AIX since Version 4.2.1. It enables continued operation of the
system, even in the case of failure on an operating system disk. Two or three
copies can be created and accessed in case of failure.
Beginning with AIX 5L Version 5.1, it is possible to designate disks as hot spare
disks in a volume group and to specify a policy to be used in the case of failing
disks. This enables fast recovery for mirrored disks when a failure occurs, using
the automatic sparing option. As soon as one mirrored disk fails, data is copied to
the spare disk or to a pool of disks. When the copy finishes, the system is again
protected against disk failures.
For detailed information on rootvg mirroring on AIX, please refer to AIX Logical
Volume Manager, From A to Z: Introduction and Concepts, SG24-5432. For LVM
hot spare disk support in a volume group, please refer to AIX 5L Differences
Guide, SG24-5765.
Chapter 5. Reliability, availability, and serviceability
207
5.7.4 TCP/IP RAS enhancements
Beginning with AIX 5L Version 5.1, AIX provides several availability features in
the TCP/IP network access. It is possible to completely eliminate failures due to
network adapters, failed gateways, and incorrect routes, using these functions.
The Dead Gateway Detection (DGD) feature in AIX 5L Version 5.1 implements a
mechanism for hosts to detect a dysfunctional gateway, adjust its routing table
accordingly, and re-route network traffic to an alternate backup route, if available.
With the multi-path routing feature in AIX 5L Version 5.1, routes no longer need
to have a different destination, netmask, or group ID lists. If there are several
routes that equally qualify as a route to a destination, AIX will use a cyclic
multiplexing mechanism (round-robin) to choose between them. The benefit of
this feature is twofold:
򐂰 Enablement of load balancing between two or more gateways.
򐂰 Feasibility of load balancing between two or more interfaces on the same
network can be realized.
These two features working together enable the system to route TCP/IP traffic
through multiple routes and avoid the defective routes. This is as important as the
server availability itself, because a network failure also causes an interrupt in
server access.
For Ethernet adapters, the network interface backup support allows multiple
adapters to be grouped together and act as a single interface. In case of a failure
in one interface, another interface will automatically take over the TCP/IP traffic
and continue operation.
For more information on how to configure this feature, please refer to Managing
AIX Server Farms, SG24-6606.
208
IBM ^ pSeries 670 and pSeries 690 System Handbook
A
Appendix A.
Minimum and default
configurations
This appendix contains the minimum configuration listed in the sales manual as
well as the e-config default configuration of pSeries 670 and pSeries 690.
© Copyright IBM Corp. 2003. All rights reserved.
209
A.1 pSeries 670 configurations
The pSeries 670 least expensive configuration contains the following features
listed in Example A-1.
Example: A-1 pSeries 670 minimum configuration
One - 7040-671 Central Electronics Complex (17U rack space required)
One - CD-ROM (#2624)
One - Diskette Drive Cable (#3155)
One - Operator Panel Attachment Cable (#3255)
One - 4 GB Memory Card, Inward Facing (#4196)
One - 4-Way, POWER4 Processor with 128 MB L3 Cache (#5253)
One - Processor Clock Card, Programmable (#5251)
One - Power Cable Group, Bulk Power to CEC and Fans (#6161)
One - Interface Cable, Power to Service Processor (#6162)
Two - Power Converter Assembly, Central Electronics Complex (#6170)
One - Power Cable Group, CEC to Power Controller, First Processor Module (#6181)
One - Capacitor Book, First and Second Processor Modules (#6198)
One - Service Processor with Remote I/O Loop Attachment, Two Loops (#6404)
One - Media Drawer, Operator Panel, Diskette (#8692) - (1U Rack Space Required)
One - Drawer Placement Indicator, 18U Position, Primary Rack (#4618)
One - Language Specify (#9xxx)
One - 7040-61D I/O Drawer (4U rack space required)
One - SCSI Cable - I/O Drawer to Media Drawer (#2122)
Two - Remote I/O Cables, 2M (#3149)
Two - 18.2 GB, 10,000 RPM Ultra3 SCSI Disk Drive Assembly (#3157)
One - I/O Drawer Attachment Cable Group, Drawer Position #4609 (#6121)
Two - Power Converter Assembly, I/O Drawer (#6172)
One - Power Cable, I/O Drawer to Media Drawer (#6179)
One - PCI Single Ended SCSI Adapter (Media Drawer Attach) (#6206)
One - I/O Drawer PCI Planar, 10 Slot, 2 Integrated Ultra3 SCSI Ports (#6563)
Two - Ultra3 SCSI 4-Pack Hot Swap Back Plane (#6564)
One - Drawer Placement Indicator, 9U Position, Primary Rack (#4609)
One - Language Specify (#9xxx)
One - 7040-61R System Rack (24-inch rack with 42U available space)
One - Front Door, Black with Copper Accent, Primary Rack (#6070)
One - Rear Door, Slim Line, Primary or Secondary Rack (#6074)
Two - Bulk Power Regulators (#6186)
Two - Power Control, Four Cooling Fans,Three DC Power Converter Connections (#6187)
Two - Power Distribution Assembly, Ten DC Power Converter Connections (#6188)
One - Bulk Power Assembly, Redundant (#8690)
One - Rack Content Specify, 7040/671 - 17U (#0209)
One - Rack Content Specify, 7040/61D - 4U (#0191)
One - Rack Content Specify, Media Drawer
One - Rack Content Specify -7040/61R Feature #8490 - 8U (#0193)
Two - Line Cords, 60A/240VAC, 6-AWG, 14 feet, IEC309 Plug (#8678)
One - Language Specify (#9xxx)
210
IBM ^ pSeries 670 and pSeries 690 System Handbook
The default configuration in Example A-2 lacks the mandatory rear door, contains
a full RIO planar while the minimum configuration contains a half populated IO
planar and contains country (or region) specific features, chosen according to
e-config country (or region) settings.
Example: A-2 pSeries 670 default configuration
Product
Description
________ __________________________________________________
7040-671
2624
3155
3255
4196
4618
4651
5005
5251
5253
6161
6162
6170
6181
6198
6404
8121
8692
9670
9703
7040-61D
2122
3145
3149
3158
4609
4651
6121
Qty
_____
Rack Server 1:pSeries 670 Model 671 CEC
32X (Max) SCSI-2 CD-ROM Drive
Diskette Drive Cable
Operator Panel Attachment Cable
4GB Memory Card, Inward Facing
Drawer Placement Indicator, 18U Position,
Primary Rack
Rack Indicator, Rack #1
Software Preinstall (RS)
Processor Clock Card, Programmable
4-Way POWER4 p670 Processor
Cable Group, Power Controller to CEC and
Fans
Interface Cable, Service Processor to Power
Subsystem
Power Converter Assembly, Central Electronics
Complex
Power Cable Group, CEC to Power Controller,
First Processor Module
Capacitor Book, Two Processor Modules
Support Processor with Remote I/O Loop Attachment,
Two Loops
Attachment Cable, Hardware Management Console
to Host, 15-Meter
Media Drawer, Operator Panel, Diskette
Manufacturing Optimization Indicator
Language Group: French
I/O Drawer 1:Model 61D I/O Drawer
SCSI Cable - I/O Drawer to Media Drawer
Remote I/O Cable, 0.5M
Remote I/O Cable, 2M
36.4 GB DISK DRIVE ASSEMBLY
DRAWER PLACEMENT IND, 9U, PRI
Rack Indicator, Rack #1
I/O Drawer Attachment Cable Group, Drawer
Position #4609
6172 Power Converter Assembly, I/O Drawer
6179 Power Cable, I/O Drawer to Media Drawer
1
1
1
1
1
1
1
1
1
1
1
1
2
1
1
1
1
1
1
1
1
1
1
2
2
1
1
1
2
1
Appendix A. Minimum and default configurations
211
6206 Ultra SCSI Adapter - SE (PCI)
6563 I/O Drawer PCI Planar, 10 Slot, 2 Integrated
Ultra3 SCSI Ports
6564 Ultra3 SCSI 4-Pack Hot Swap Back Plane
9703 Language Group: French
7040-61R 7040-61R : Rack 1:Model 61R Rack
0191 Rack Content Specify: 7040/61D - 4U
0192 Rack Content Specify: Media Drawer Feature
#8692 - 1U
0193 Rack Content Specify: 7040/61R Feature #8690
- 8U
0209 Rack Content Specify: 7040/671 - 17U
4651 Rack Indicator, Rack #1
6070 Front Door, Black with Copper Accent, Primary
Rack
6186 Bulk Power Regulator
6187 Power Controller, Four Cooling Fans, Three
DC Power Converter Connections
6188 Power Distribution Assembly
8677 Line Cord, 8AWG, 14ft, No Plug
8690 Bulk Power Assembly, Redundant
9703 Language Group: French
1
2
4
1
1
1
1
1
1
1
1
2
2
2
2
1
1
A.2 pSeries 690 configurations
The pSeries 690 least expensive configuration contains the features listed in
Example A-3.
Example: A-3 pSeries 690 minimum configuration
One - 7040-681 Central Electronics Complex (17U rack space required):
One - CD-ROM (#2624)
One - Diskette Drive Cable (#3155)
One - Operator Panel Attachment Cable (#3255)
One - 128 MB Level 3 Cache (4 X 32 MB), 400 MHz (#4138)
One - 8 GB Memory Card, Inward Facing (#4181)
One - 8-Way, POWER4 Processor (First) (#5242)
One - Processor Clock Card, Programmable (#5251)
One - Power Cable Group, Bulk Power to CEC and Fans (#6161)
One - Interface Cable, Power to Service Processor (#6162)
Two - Power Converter Assembly, Central Electronics Complex (#6170)
One - Power Cable Group, CEC to Power Controller, First Processor Module (#6181)
One - Capacitor Book, First and Second Processor Modules (#6198)
One - Service Processor with Remote I/O Loop Attachment, Two Loops (#6404)
212
IBM ^ pSeries 670 and pSeries 690 System Handbook
One - Backplane, Central Electronics Complex (#6565)
One - Media Drawer, Operator Panel, Diskette (#8692) - (1U Rack Space Required)
One - Drawer Placement Indicator, 18U Position, Primary Rack (#4618)
One - Language Specify (#9xxx)
One - 7040-61D I/O Drawer (4U rack space required):
One - SCSI Cable - I/O Drawer to Media Drawer (#2122)
One - Remote I/O Cable, 05.M (#3145) (between drawer halves)
Two - Remote I/O Cables, 2M (#3149)
Two - 18.2 GB, 10,000 RPM Ultra3 SCSI Disk Drive Assembly (#3157)
One - I/O Drawer Attachment Cable Group, Drawer Position #4605 (#6121)
Two - Power Converter Assembly, I/O Drawer (#6172)
One - Power Cable, I/O Drawer to Media Drawer (#6179)
One - PCI Single Ended SCSI Adapter (Media Drawer Attach) (#6206)
Two - I/O Drawer PCI Planar, 10 Slot, 2 Integrated Ultra3 SCSI Ports (#6563)
Four - Ultra3 SCSI 4-Pack Hot Swap Back Plane (#6564)
One - Drawer Placement Indicator, 9U Position, Primary Rack (#4609)
One - Language Specify (#9xxx)
One - 7040-61R System Rack (24-inch rack with 42U available space):
One - Front Door, Black with Copper Accent, Primary Rack (#6070)
One - Rear Door, Slim Line, Primary or Secondary Rack (#6074)
Two - Bulk Power Regulators (#6186)
Two - Power Control,Four Cooling Fans, Three DC Power Converter Connections (#6187)
Two - Power Distribution Assembly, Ten DC Power Converter Connections (#6188)
One - Bulk Power Assembly, Redundant (#8690)
One - Rack Content Specify, 7040/681 - 17U (#0190)
One - Rack Content Specify, 7040/61D - 4U (#0191)
One - Rack Content Specify, 7040/681 Feature #8692 - 1U (#0192)
One - Rack Content Specify -7040/61R Feature #8490 - 8U (#0193)
Two - Line Cords, 60A/240VAC, 6-AWG, 14 feet, IEC309 Plug (#8678)
One - Language Specify (#9xxx)
The default configuration in Example A-4 lacks the mandatory rear door, and
contains country (or region) specific features, chosen according to e-config
country (or region) settings.
Example: A-4 pSeries 690 default configuration
Product
________
7040-681
2624
3155
3255
4138
4181
4618
Description
__________________________________________________
Rack Server 1:pSeries 690 Model 681 CEC
32X (Max) SCSI-2 CD-ROM Drive
Diskette Drive Cable
Operator Panel Attachment Cable
128MB Level 3 Cache (4 X 32MB), 400MHz
8GB Memory Card, Inward Facing
Drawer Placement Indicator, 18U Position,
Primary Rack
4651 Rack Indicator, Rack #1
5005 Software Preinstall (RS)
Qty
_____
1
1
1
1
1
1
1
1
1
Appendix A. Minimum and default configurations
213
5242 8-Way POWER4 Processor
5251 Processor Clock Card, Programmable
6161 Cable Group, Power Controller to CEC and
Fans
6162 Interface Cable, Service Processor to Power
Subsystem
6170 Power Converter Assembly, Central Electronics
Complex
6181 Power Cable Group, CEC to Power Controller,
First Processor Module
6198 Capacitor Book, Two Processor Modules
6404 Support Processor with Remote I/O Loop Attachment,
Two Loops
6565 Backplane, Central Electronics Complex
8121 Attachment Cable, Hardware Management Console
to Host, 15-Meter
8692 Media Drawer, Operator Panel, Diskette
9703 Language Group: French
7040-61D
2122
3145
3149
3158
4609
4651
6121
6172
6179
6206
6563
6564
9703
7040-61R
0190
0191
0192
0193
4651
6070
6186
6187
214
I/O Drawer 1:Model 61D I/O Drawer
SCSI Cable - I/O Drawer to Media Drawer
Remote I/O Cable, 0.5M
Remote I/O Cable, 2M
36.4 GB DISK DRIVE ASSEMBLY
Drawer Placement Indicator, 9U Position,
Primary Rack
Rack Indicator, Rack #1
I/O Drawer Attachment Cable Group, Drawer
Position #4609
Power Converter Assembly, I/O Drawer
Power Cable, I/O Drawer to Media Drawer
Ultra SCSI Adapter - SE (PCI)
I/O Drawer PCI Planar, 10 Slot, 2 Integrated
Ultra3 SCSI Ports
Ultra3 SCSI 4-Pack Hot Swap Back Plane
Language Group: French
7040-61R : Rack 1:Model 61R Rack
Rack Content Specify: 7040/681 - 17U
Rack Content Specify: 7040/61D - 4U
Rack Content Specify: Media Drawer Feature
#8692 - 1U
Rack Content Specify: 7040/61R Feature #8690
- 8U
Rack Indicator, Rack #1
Front Door, Black with Copper Accent, Primary
Rack
Bulk Power Regulator
Power Controller, Four Cooling Fans, Three
IBM ^ pSeries 670 and pSeries 690 System Handbook
1
1
1
1
2
1
1
1
1
1
1
1
1
1
1
2
2
1
1
1
2
1
1
2
4
1
1
1
1
1
1
1
1
2
2
6188
8677
8690
9703
DC Power Converter Connections
Power Distribution Assembly
Line Cord, 8AWG, 14ft, No Plug
Bulk Power Assembly, Redundant
Language Group: French
2
2
1
1
Appendix A. Minimum and default configurations
215
216
IBM ^ pSeries 670 and pSeries 690 System Handbook
B
Appendix B.
I/O loop cabling and
performance
This section first presents five graphical views of loop cabling, using either RIO
components, RIO-2 components, or a mix of the two technologies. The diagrams
contains an indication of the data throughput burst and a sustained rate that can
be expected in each part of the loops.
The three following figures present examples of the cabling of fully configured
pSeries 690 with the maximum possible number of installed IO drawers.
© Copyright IBM Corp. 2003. All rights reserved.
217
p690 RIO IO drawer
RIO planars with single RIO loop
Burst
1700MB/s Simplex
3400MB/s Duplex
Sustained
1250MB/s Simplex
2500MB/s Duplex
GX bus @ 433MHz
for 1.3 GHz system
2 Planar per RIO HUB
RIO HUB
Per RIO link
Burst
500MB/s Simplex
1000MB/s Duplex
Sustained
400MB/s Simplex
550MB/s Duplex
RIO
RIO
Active
RIO
64 bit PHBs
500MB/s Burst
300MB/s Sustained
64
RIO
Link 0
Active
Passive/failover
Link 0
RIO to PCI
Bridge
64
32
PHB2
PCI to PCI
PHB3
PCI to PCI
Bridge
Bridge
Bridge
64
33
66 = 64 bit 66MHz PCI 3.3V signaling
Link 1
RIO to PCI
Bridge
PHB1
PCI to PCI
Ultra3
SCSI
RIO
RIO
Link 1
66
Planar 1
FC 6563
Per Planar
Burst
500MB/s Simplex
1000MB/s Duplex
Sustained
400MB/s Simplex
550MB/s Duplex
Ultra3
SCSI
64
PHB1
PCI to PCI
PHB2
PCI to PCI
Bridge
Bridge
32 bit PHB
250MB/s Burst
150MB/s Sustained
32
PHB3
PCI to PCI
Bridge
66
Planar 2
FC 6563
33
Ultra3
SCSI
Ultra3
SCSI
7040-61D I/O drawer
33 = 64 bit 33MHz PCI 5.0V signaling
Figure B-1 RIO drawer, in single-loop mode on a 1.3 GHz server
In Figure B-1, all components are using the RIO technology. The two planers are
connected to the same loop. Each planar therefore transfers data through only
one RIO link.
The transfer rate within the CEC between the RIO book (RIO HUB) and the MCM
depends on the processor speed.
218
IBM ^ pSeries 670 and pSeries 690 System Handbook
p690 RIO-2 IO drawer
RIO-2 planars with dual RIO-2 Loops
Burst
2250MB/s Simplex
4500MB/s Duplex
Sustained
1700MB/s Simplex
3400MB/s Duplex
GX bus @ 566MHz
for 1.7 GHz system
1 Planar per RIO2 HUB
RIO2
HUB
Per RIO2 link
Burst
1000MB/s Simplex
2000MB/s Duplex
Sustained
750MB/s Simplex
1000MB/s Duplex
RIO2
HUB
RIO 2
RIO 2
RIO 2
RIO 2
RIO 2
RIO 2
Per Planar
Burst
2000MB/s Simplex
3000MB/s Duplex
Sustained
1350MB/s Simplex
1600MB/s Duplex
All links active
RIO 2
RIO 2
64 bit PHBs
Link 0
Link 1
1000MB/s Burst
RIO2 to PCI-X
600MB/s Sustained
Link 0
Link 1
RIO2 to PCI-X
Bridge
Bridge
PHB1
PCI-X to PCI-X
PHB2
PCI-X to PCI-X
PHB3
PCI-X to PCI-X
PHB1
PCI-X to PCI-X
PHB2
PCI-X to PCI-X
Bridge
Bridge
PHB3
PCI-X to PCI-X
Bridge
Bridge
Bridge
Bridge
Planar 1
FC 6571
PHB 1 routed via RIO2 link 0
Ultra3
SCSI
Ultra3
SCSI
Planar 2
FC 6571
7040-61D I/O drawer
Ultra3
SCSI
Ultra3
SCSI
PHBs 2 and 3 routed via RIO2 link 1
All slots are PCI-X 64 bit 133 MHz 3.3V signaling also support PCI 3.3V signaling atapters
Figure B-2 RIO-2 drawer, in dual-loop mode on a 1.7GHz server
In Figure B-2, all components are using the RIO-2 technology. Each of the two
planers are using the two links of a dedicated loop.
This is the recommended and default loop configuration for RIO-2 loops.
Appendix B. I/O loop cabling and performance
219
p690 RIO-2 IO drawer
RIO-2 planars with single RIO-2 Loop
Burst
2250MB/s Simplex
4500MB/s Duplex
Sustained
1700MB/s Simplex
3400MB/s Duplex
GX bus @ 566MHz
for 1.7 GHz system
2 Planar per RIO2 HUB
Per Planar
Burst
1000MB/s Simplex
2000MB/s Duplex
Sustained
750MB/s Simplex
1000MB/s Duplex
RIO2
HUB
Per RIO2 link
Burst
1000MB/s Simplex
2000MB/s Duplex
Sustained
750MB/s Simplex
1000MB/s Duplex
RIO 2
RIO 2
Active
RIO 2
RIO 2
Active
Passive/failover
RIO 2
RIO 2
Link 0
Link 1
64 bit PHBs
Link 0
Link 1
1000MB/s Burst
RIO2 to PCI-X
600MB/s Sustained
RIO2 to PCI-X
Bridge
Bridge
PHB1
PCI-X to PCI-X
PHB2
PCI-X to PCI-X
PHB3
PCI-X to PCI-X
PHB1
PCI-X to PCI-X
PHB2
PCI-X to PCI-X
Bridge
Bridge
PHB3
PCI-X to PCI-X
Bridge
Bridge
Bridge
Bridge
Planar 1
FC 6571
Ultra3
SCSI
Ultra3
SCSI
Planar 2
FC 6571
Ultra3
SCSI
Ultra3
SCSI
7040-61D I/O drawer
Figure B-3 RIO-2 drawer, in single-loop mode on a 1.7 GHz server
In Figure B-3, all components are using the RIO-2 technology, but the two
planers are connected in a single-loop mode, as would be an RIO loop. Each
planar therefore transfers data through only one RIO link.
This does not yield the best performance available using these components.
However, the throughput of this RIO-2 loop is 32% higher than for the RIO loop of
Figure B-1.
220
IBM ^ pSeries 670 and pSeries 690 System Handbook
p690 Mixed IO drawer
Mixed IO planars with dual RIO2/RIO Loops
Burst
2250MB/s Simplex
4500MB/s Duplex
Sustained
1700MB/s Simplex
3400MB/s Duplex
GX bus @ 566MHz
for 1.7 GHz system
1 Planar per RIO2 HUB
RIO 2
HUB
RIO 2
HUB
RIO 2
RIO 2
Both links active
RIO 2
Link 0
RIO 2
RIO 2
One link active
One link for failover
RIO 2
RIO2 to RIO cables
RIO
RIO
Link 0
Link 1
Link 1
RIO to PCI
Bridge
RIO2 to PCI-X
Bridge
64
PHB1
PCI-X to PCI-X
PHB2
PCI-X to PCI-X
PHB3
PCI-X to PCI-X
Bridge
Bridge
Bridge
64
PHB1
PCI to PCI
PHB2
PCI to PCI
Bridge
Bridge
32
PHB3
PCI to PCI
Bridge
66
Planar 1
FC 6571
Ultra3
SCSI
Ultra3
SCSI
Planar 2
FC 6563
33
Ultra3
SCSI
Ultra3
SCSI
7040-61D I/O drawer
Figure B-4 Mixed RIO/RIO-2 drawer, in dual-loop mode on a 1.7 GHz server
In Figure B-4, one of the I/O planars in the drawer is using the RIO technology,
therefore providing three I/O slots for PCI 5V adapters. Each of the two planers
are connected in a single-loop mode, as would be an RIO loop. However, only
the RIO-2 planar transfers data through to the links, while the RIO planar can fly
transmit data through one RIO link.
Appendix B. I/O loop cabling and performance
221
p690 RIO IO drawer
Migrated RIO drawer on RIO2 book with single RIO loop
GX bus @ 566MHz
for 1.7 GHz system
2 Planar per RIO HUB
RIO 2
HUB
RIO 2
RIO
64 bit PHBs
500MB/s Burst
300MB/s Sustained
RIO 2
RIO2 to RIO cables
Active
Active
RIO
Link 0
Link 0
RIO to PCI
Bridge
32
PHB1
PCI to PCI
PHB2
PCI to PCI
PHB3
PCI to PCI
Bridge
Bridge
Bridge
64
33
Ultra3
SCSI
66 = 64 bit 66MHz PCI 3.3V signaling
Link 1
RIO to PCI
Bridge
66
Planar 1
FC 6563
RIO2
RIO
Link 1
64
64
Passive/failover
Ultra3
SCSI
64
PHB1
PCI to PCI
PHB2
PCI to PCI
Bridge
Bridge
32 bit PHB
250MB/s Burst
150MB/s Sustained
32
PHB3
PCI to PCI
Bridge
66
Planar 2
FC 6563
33
Ultra3
SCSI
Ultra3
SCSI
7040-61D I/O drawer
33 = 64 bit 33MHz PCI 5.0V signaling
Figure B-5 Migrated single-loop mode on a 1.7 GHz server
Figure B-5, presents a configuration that can be found in a pSeries 690 where an
MES has been installed, replacing the processors and RIO books by 1.7 GHz
MCM and RIO-2 books.
The configuration of the RIO loop is unchanged, identical to the one presented in
Figure B-1. However, the RIO cables have been replaced with RIO to RIO-2
cables.
222
IBM ^ pSeries 670 and pSeries 690 System Handbook
p690+ I/O Support
Max bandwidht per PCI slot All drawers double loop 7 drawers max
MCM0
000 00 (s); 0A
000 11 (v); 0C
Base IO
1
0
Book
1
0
P0
P1
GX-RIO2P2
P3
0
12
P0
B0
GX-RIO2
P1
1
000 10 (u); 0D
000 01 (t); 0B
B1
Flex
RIO-PCI
Service
PCI
Processor
I/O Slot 0
I/O Slot 1
IO expansion
2
9
P0
GX-RIO2P1
8
MCM1
8
001 11 (v); 1A
001 00 (s); 1C
1
P0
GX-RIO2
P1
9
001 01 (t); 1D
10
11
2
P0
GX-RIO2P1
10
MCM2
8
010 00 (s); 2A
010 11 (v); 2C
1
P0
GX-RIO2
P1
11
010 10 (u); 2D
10
11
2
P0
1
P0
9
011 00 (s); 3C
1
0
I/O Drawer3
Primary Rack EIA Pos 3
1
0
1
0
I/O Drawer 2
Primary Rack EIA Pos 5
1
0
1
0
I/O Drawer 5 *
Expansion Rack EIA Pos 1
1
0
1
0
B0
2
P0
1
P0
P0
1
P0
GX-RIO2
P1
7
10
1
0
B0
B1
2
A0
A1
B0
B1
A0
A1
B0
B1
2
P0
GX-RIO2
P1
12
11
I/O Drawer 6 *
Expansion Rack EIA Pos 5
B1
011 01 (t); 3D
011 10 (u); 3B
1
0
A1
GX-RIO2P1
6
8
1
0
A1
IO expansion
MCM3
011 11(v): 3A
A0
A0
GX-RIO2
P1
5
I/O Slot 3
I/O Drawer 7 *
Expansion Rack EIA Pos 9
B0
GX-RIO2P1
4
010 01 (t); 2B
1
0
B1
IO expansion
9
1
0
A1
GX-RIO2
P1
3
I/O Slot 2
HSC
Op-Pannel
Diskette
A0
GX-RIO2P1
2
001 10 (u);1B
I/O Drawer 1
Primary Rack EIA Pos 9
1
0
A0
A1
1
P0
GX-RIO2P1
13
A0
I/O Drawer 4 *
Primary Rack EIA Pos 1
A1
B0
B1
* Drawer positions may vary due to IBF options
Figure B-6 Configuration for maximum bandwidth
In an I/O bound system where I/O adapters and internal disks access are the
most critical bottleneck, you can obtain the maximum bandwidth on each adapter
by having as many adapters as possible, and the smallest number of adapters for
each uplink between the IO books and the MCM. This is provided by a seven IO
drawers configuration, all connected in dual-loop mode (Figure B-6).
Appendix B. I/O loop cabling and performance
223
p690+ I/O Support
8 drawers max system bandwidht 6 drawers double loop and 2 drawers single loop
MCM0
000 00 (s); 0A
000 11 (v); 0C
Base IO
1
0
Book
1
0
P0
P1
GX-RIO2P2
P3
0
12
P0
B0
GX-RIO2
P1
1
000 10 (u); 0D
000 01 (t); 0B
B1
Flex
RIO-PCI
Service
PCI
Processor
I/O Slot 0
I/O Slot 1
IO expansion
2
9
P0
GX-RIO2P1
8
MCM1
8
1
P0
GX-RIO2
P1
9
001 11 (v); 1A
001 00 (s); 1C
001 01 (t); 1D
10
11
A1
B0
2
P0
1
P0
2
P0
GX-RIO2P1
10
MCM2
8
010 00 (s); 2A
010 11 (v); 2C
1
P0
GX-RIO2
P1
11
010 10 (u); 2D
10
010 01 (t); 2B
11
011 00 (s); 3C
8
1
2
P0
1
P0
P0
1
0
I/O Drawer 6 *
Expansion Rack EIA Pos 5
1
0
1
0
I/O Drawer3
Primary Rack EIA Pos 3
1
0
1
0
I/O Drawer 2
Primary Rack EIA Pos 5
1
0
1
0
1
0
1
0
B0
A0
A1
B0
B1
P0
A0
2
1
I/O Drawer 5 *
Expansion Rack EIA Pos 1
A1
B0
B1
P0
GX-RIO2P1
12
11
1
0
B1
011 01 (t); 3D
011 10 (u); 3B
I/O Drawer 8*
Expansion Rack EIA Pos 13
B0
GX-RIO2
P1
7
10
1
0
B1
GX-RIO2P1
6
MCM3
011 11(v): 3A
1
0
A1
IO expansion
2
I/O Drawer 7*
Expansion Rack EIA Pos 9
A1
GX-RIO2P1
5
9
1
0
A0
GX-RIO2P1
4
I/O Slot 3
A0
IO expansion
9
1
0
B1
GX-RIO2P1
3
I/O Slot 2
HSC
Op-Pannel
Diskette
A0
GX-RIO2P1
2
001 10 (u);1B
I/O Drawer 1
Primary Rack EIA Pos 9
1
0
A0
A1
P0
GX-RIO2P1
13
A0
I/O Drawer 4 *
Primary Rack EIA Pos 1
A1
B0
B1
* Drawer positions may vary due to IBF options
Figure B-7 Configuration for maximum number of adapters and disks
If you are looking for as many adapters and disks as possible in your
configuration (for example, in a server consolidation environment with 32
partitions and many adapters per LPAR), and you also want to obtain the best
possible I/O throughput, you need 8 I/O drawers, but since there are only 14 pairs
of RIO-2 ports in the IO books, you need to configure two drawers in single-loop
mode, and the six others in dual-loop mode as in Figure B-7.
224
IBM ^ pSeries 670 and pSeries 690 System Handbook
p690 I/O Support
8 drawers max Migration compatibility
MCM0
000 00 (s); 0A
000 11 (v); 0C
000 10 (u); 0D
Base IO Book
P0
P1
GX-RIO2P2
P3
0
0
1
12
A0
A1
P0
B0
P1
B1
GX-RIO2
1
000 01 (t); 0B
1
0
1
0
1
0
1
0
I/O Drawer 1
Primary Rack EIA Pos 9
I/O Drawer 2
Primary Rack EIA Pos 5
Flex
RIO-PCI
HSC
Op-Pannel
Diskette
Service
Processor
PCI
I/O Slot 0
I/O Slot 1
MCM1
001 11 (v); 1A
001 00 (s); 1C
001 01 (t); 1D
001 10 (u);1B
I/O Slot 2
IO expansion
MCM2
010 11 (v); 2C
P0
GX-RIO2P1
10
8
010 00 (s); 2A
2
9
1
P0
GX-RIO2
P1
11
010 10 (u); 2D
10
11
A1
B0
011 11(v): 3A
2
P0
1
P0
A0
1
0
I/O Drawer 6
Expansion Rack EIA Pos 5
1
0
1
0
A1
B0
B1
9
2
P0
1
P0
GX-RIO2P1
6
1
0
1
0
8
GX-RIO2
P1
7
A0
10
2
1
I/O Drawer 4
Primary Rack EIA Pos 13 or
Expansion Rack EIA Pos 9
B0
B1
P0
GX-RIO2
P1
12
11
I/O Drawer 3
Primary Rack EIA Pos 1
A1
1
0
011 01 (t); 3D
011 10 (u); 3B
1
0
IO expansion
MCM3
011 00 (s); 3C
I/O Drawer 5
Expansion Rack EIA Pos 1
B1
GX-RIO2
P1
5
I/O Slot 3
1
0
A0
GX-RIO2P1
4
010 01 (t); 2B
1
0
P0
GX-RIO2P1
13
1
0
A0
A1
B0
B1
1
0
1
0
I/O Drawer 7
Expansion Rack EIA Pos 13
I/O Drawer 8
Expansion Rack EIA Pos 19
Figure B-8 Configuration after migration to RIO-2 Books without recabling.
Figure B-8 shows a valid configuration after migrating from RIO to RIO-2 books.
The previous cabling topology has been kept, to maintain the LPAR
configurations. However, the RIO cables have been replaced by RIO to RIO-2
cables.
Appendix B. I/O loop cabling and performance
225
226
IBM ^ pSeries 670 and pSeries 690 System Handbook
Abbreviations and acronyms
AC
Alternating Current
DCA
AIX
Advanced Interactive
eXecutive
Distributed Converter
Assembly
DDR
Double Data Rate
APAR
Authorized Program Analysis
Report
DGD
Dead Gateway Detection
DIMM
Dual Inline Memory Module
ATM
Asynchronous Transfer Mode
DLPAR
Dynamic logical partitioning
BI
Business Intelligence
DMA
Direct Memory Access
BIST
Built-in Self Test
DMTF
BPA
Bulk Power Assembly
Distributed Management Task
Force
BPC
Bulk Power Controller
DRAM
BPF
Bulk Power Fan
Dynamic Random Access
Memory
BPR
Bulk Power Regulator
DVD
Digital Versatile Disk
bps
Bits per Second
DVD-RAM
CD
Compact Disk
Digital Versatile Disk Random Access Media
CD-ROM
Compact Disk - Read Only
Media
DVD-ROM
Digital Versatile Disk Random Only Media
CEC
Central Electronics Complex
ECC
Error Checking and
Correcting
cfm
Cubic feet per minute
e-config
CIM
Common Interface Model
IBM Configurator for
e-business
CIU
Core Interface Unit
EEH
Extended Error Handling
CMOS
Complementary
Metal-Oxide-Silicon
EIA
Electronic Industry Alliance
EMF
Electro-motive Force
CPU
Central Processing Unit
EPO
Emergency Power On
CRAD
Customer Requested Arrival
Date
EPOW
Early Power Off Warning
ERP
Enterprise Resource Planning
ESCON
Enterprise Systems
Connection
FC
Feature Code
FDDI
Fibre Distributed Data
Interface
FFDC
First Failure Data Capture
FIR
Fault Isolation Register
FLOP
Floating-point Operation
CRC
Cyclic Redundancy Check
CRM
Customer Relationship
Management
CTA
Click to Accept
CUoD
Capacity Upgrade on
Demand
DAT
Digital Audio Tape
DC
Direct Current
© Copyright IBM Corp. 2003. All rights reserved.
227
FQDN
Fully Qualified Domain Name
LPAR
Logical Partition
FRU
Field Replaceable Unit
LVD
Low Voltage Differential
ft
feet
LVM
Logical Volume Manager
GA
General Announcement
MA
Maintenance Agreement
GB
Gigabyte
MB
Megabyte
GHz
Gigahertz
Mbps
Megabit per Second
HACMP
High Availability Cluster
Multiprocessing
MCM
Multichip Module
MDA
Motor Drive Assembly
HAEH
High Availability Event
Handler
MES
Miscellaneous Equipment
Specification
HIPPI
High Performance Parallel
Interface
MHz
Megahertz
HMC
IBM Hardware Management
Console for pSeries
mm
millimeter
MPOA
Multiprotocol over ATM
MRPD
Machine ReportedProduct
Data
MSA
Motor/Scroll Assembly
NC
Non-cacheable
NIM
Network Installation
Management
NIS
Network Information Service
NVRAM
Non-volatile Random Access
Memory
HPC
High Performance Computing
I/O
Input/Output
IBF
internal battery feature
IBM
International Business
Machines Corporation
ID
identification
IDE
Integrated Data Equipment
IP
Internet Protocol
IPL
Initial Program Load
ODM
Object Data Manager
ISDN
Integrated Services Digital
Network
OLAP
Online Analytical Processing
ISV
Independent Software Vendor
PC
Personal Computer
ITSO
International Technical
Support Organization
PCI
Peripheral Component
Interconnect
JTAG
Joint Test Action Group
PFT
Page Frame Table
kW
kilowatt
PHB
PCI Host Bridge
kwh
Kilowatt per Hour
POR
Power On Reset
L1
Level 1
POST
Power-On Self-Test
L2
Level 2
POWER
Performance Optimization
With Enhanced RISC
L3
Level 3
PSSP
LAN
Local Area Network
Parallel System Support
Program
LDAP
Lightweight Directory Access
Protocol
QBB
Quad Building Block
RAM
Random Access Memory
LED
Light Emitting Diode
228
IBM ^ pSeries 670 and pSeries 690 System Handbook
RAN
Remote Asynchronous Node
UTP
Untwisted pair
RAS
Reliability, Availability, and
Serviceability
VHDCI
Very High Density Connector
Interface
RIO
Remote I/O
VM
Virtual Machine
RISC
Reduced Instruction Set
Computer
VPD
Vital Product Data
WAN
Wide Area Network
RMC
Resource Monitoring and
Control
RML
Real Mode Limit
RMO
Real Mode Offset
RPM
Rotation per Minute
RTAS
Run Time Abstraction
Services
SBE
Single Bit Error
SCM
Supply Chain Management
SCSI
Small Computer System
Interface
SDRAM
Synchronous DRAM
SES
SCSI Enclosure Service
SMI
Synchronous Memory
Interface
SMIT
System Management
Interface Tool
SMP
Symmetric Multiprocessing
SMS
System Management
Services
SOI
Silicon-on-Insulator
SP
IBM RS/6000 Scalable
POWERparallel® Systems
SSA
Serial Storage Architecture
TB
Terabyte
TCE
Translation Control Entry
TCP
Transmission Control Protocol
TLB
Translation Look-aside Buffer
TOD
Time-of-Day
UPS
Uninterruptible Power Supply
URL
Uniform Resource Locator
USB
Universal Serial Bus
Abbreviations and acronyms
229
230
IBM ^ pSeries 670 and pSeries 690 System Handbook
Related publications
The publications listed in this section are considered particularly suitable for a
more detailed discussion of the topics covered in this redbook.
IBM Redbooks
For information on ordering these publications, see “How to get IBM Redbooks”
on page 234. Note that some of the documents referenced here may be available
in softcopy only.
򐂰 AIX 5L Differences Guide, SG24-5765
򐂰 AIX Logical Volume Manager, From A to Z: Introduction and Concepts,
SG24-5432
򐂰 A Practical Guide for Resource Monitoring and Control (RMC), SG24-6606
򐂰 Managing AIX Server Farms, SG24-6606
򐂰 The Complete Partitioning Guide for IBM ^ pSeries Servers,
SG24-7039
򐂰 The POWER4 Processor Introduction and Tuning Guide, SG24-7041
Other publications
These publications are also relevant as further information sources. (The
publications marked with an asterisk (*) are located on the documentation
CD-ROM that ships with the AIX operating system).
򐂰 AIX 5L Version 5.2 AIX Installation in a Partitioned Environment *
򐂰 AIX 5L Version 5.2 Installation Guide and Reference *
򐂰 AIX 5L Version 5.2 Reference Documentation: Commands Reference *
򐂰 AIX 5L Version 5.2 System Management Guide: AIX 5L Version 5.2
Web-based System Manager Administration Guide *
򐂰 AIX 5L Version 5.2 System Management Guide: Communications and
Networks *
򐂰 AIX 5L Version 5.2 System Management Guide: Operating System and
Devices *
© Copyright IBM Corp. 2003. All rights reserved.
231
򐂰 Electronic Service Agent for pSeries and RS/6000 User’s Guide, available at:
ftp://ftp.software.ibm.com/aix/service_agent_code/AIX/svcUG.pdf
򐂰 Electronic Service Agent for pSeries Hardware Management Console User’s
Guide, available at:
ftp://ftp.software.ibm.com/aix/service_agent_code/HMC/HMCSAUG.pdf
򐂰 IBM ^ Cluster 1600 Harware Planning, Installation and Service,
GA22-7863
򐂰 IBM ^ pSeries 670 Installation Guide, SA38-0613-01
򐂰 IBM ^ pSeries 670 Service Guide, SA38-0615
򐂰 IBM ^ pSeries 670 User’s Guide, SA38-0614-01
򐂰 IBM ^ pSeries 690 Installation Guide, SA38-0587
򐂰 IBM ^ pSeries 690 Service Guide, SA38-0589
򐂰 IBM ^ pSeries 690 User’s Guide, SA38-0588
򐂰 IBM Hardware Management Console for pSeries Maintenance Guide,
SA38-0603
򐂰 IBM Hardware Management Console for pSeries Installation and Operations
Guide, SA38-0590
򐂰 IBM Reliable Scalable Cluster Technology for AIX 5L RSCT Guide and
Reference, SA22-7889
򐂰 IBM Reliable Scalable Cluster Technology for AIX 5L Messages, GA22-7891
򐂰 IBM Reliable Scalable Cluster Technology for AIX 5L Technical Reference,
SA22-7890
򐂰 Installation Guide 61D I/O drawer 61R Second I/O Rack, SA23-1281
򐂰 Exploring Expect: A Tcl-based Toolkit for Automating Interactive Programs, by
Don Libes. O’Reilly & Associates, Inc., January 1995. ISBN 1565920902
򐂰 Adapter, Devices, and Cable Information for Multiple Bus Systems,
SA38-0516
򐂰 RS/6000 and eServer pSeries Diagnostics Information for Multiple Bus
Systems, SA38-0509
򐂰 PCI Adapter Placement References, SA38-0538
򐂰 Site and Hardware Planning Information, SA38-0508
You can access all of the pSeries hardware related documentation through the
Internet at the following URL:
http://publib16.boulder.ibm.com/pseries/en_US/infocenter/base/hardware.htm
232
IBM ^ pSeries 670 and pSeries 690 System Handbook
You can also access all of the AIX documentation through the Internet at the
following URL:
http://publib16.boulder.ibm.com/pseries/en_US/infocenter/base/aix.htm
The following whitepapers can be found on the Internet:
򐂰 IBM ^ pSeries 690 Availability Best Practices white paper
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_avail.html
򐂰 IBM ^ pSeries 690 Configuring for Performance white paper
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_config.html
򐂰 IBM ^ pSeries 690: Reliability, Availability, Serviceability white paper
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_ras.html
򐂰 IBM ^ pSeries 690 with the HPC feature white paper
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/p690_hpc.html
򐂰 Linux for IBM ^ pSeries: An overview for customers white paper
http://www.ibm.com/servers/eserver/pseries/linux/whitepapers/linux_pseries.html
򐂰 Partitioning for the IBM ^ pSeries 690 System white paper
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/lpar.html
򐂰 POWER4 System Microarchitecture white paper
http://www.ibm.com/servers/eserver/pseries/hardware/whitepapers/power4.html
Online resources
These Web sites are also relevant as further information sources:
򐂰 AIX 5L operating system and related IBM products information
http://www.ibm.com/servers/aix/
򐂰 AIX toolkit for Linux applications
http://www.ibm.com/servers/aix/products/aixos/linux/download.html
򐂰 Application availability on the AIX 5L operating system (alphabetical listing
and advanced search options for IBM software products and third-party
software products)
http://www.ibm.com/servers/aix/products/
򐂰 CuOD process, brief explanation
http://www.ibm.com/servers/eserver/pseries/cuod/tool.html
򐂰 IBM AIX 5L Solution Developer Application Availability Web page
http://www.ibm.com/servers/aix/isv/availability.html
Related publications
233
򐂰 IBM AIX: IBM Application Availability Guide Web page
http://www.ibm.com/servers/aix/products/ibmsw/list
򐂰 IBM Configurator for e-business
http://ftp.ibmlink.ibm.com/econfig/announce/index.htm
򐂰 IBM ^ pSeries LPAR documentation and references Web site
http://www.ibm.com/servers/eserver/pseries/lpar/resources.html
򐂰 Linux for pSeries system guide
http://www.ibm.com/servers/eserver/pseries/linux
򐂰 Linux on pSeries information
http://www.ibm.com/servers/eserver/pseries/linux/
򐂰 Microcode Discovery Service information
http://techsupport.services.ibm.com/server/aix.invscoutMDS
򐂰 OpenSSH Web site
http://www.openssh.com
򐂰 VPD Capture Service
http://techsupport.services.ibm.com/server/aix.invscoutVPD
How to get IBM Redbooks
You can search for, view, or download Redbooks, Redpapers, Hints and Tips,
draft publications and Additional materials, as well as order hardcopy Redbooks
or CD-ROMs at this Web site:
ibm.com/redbooks
234
IBM ^ pSeries 670 and pSeries 690 System Handbook
Index
Symbols
/etc/hosts 203
/etc/inittab 207
/etc/netsvc.conf 203
Numerics
10/2002 system microcode update 10
128-port adapter 94
4 way 1.1GHz POWER4 10
4 way 1.3GHz POWER4 HPC 10
42 EIA height rack 18
7040-61D I/O drawer 83
7040-61R rack 92
7040-61R system rack 82–83, 92
7040-671 Central Electronics Complex 82–83
7040-681 Central Electronics Complex 82–83, 85
8 way 1.1GHz POWER4 10
8 way 1.3GHz POWER4 Turbo 10
A
AC 73
acoustic rear door 92
activation code 137
Adapter Cards tab 104
e-config 107, 110
Additional tab 104
Air Moving Device 75
AIX 5.2 13
AIX 5L Version 5.1 3
AIX 5L Version 5.2 3
AIX 5L Version 5.2 Web-based System Manager
77
AIX system error log 160
aluminum interconnect 4
AS/400 7
automatic configuration 194
automatic sparing option 207
availability 10, 157
B
base card 47–48
BI 11
© Copyright IBM Corp. 2003. All rights reserved.
BIST 26, 162
bit scattering 165
bit steering 165
blind swap cassette 174
book packaging 47
boot sequence 162
BPA 56, 73, 83
BPC 56, 73
BPD 73
BPR 73
building block 8
built-in self test 26, 162
Bulk Power Assembly 56, 73, 83
Bulk Power Controller 56, 73
Bulk Power Distributor 73
Bulk Power Regulator 73, 93
Business Intelligence 11
business-critical applications 3
C
cabling pattern 58
cache line delete 172
capacitor book 20
Capacity Upgrade on Demand 9, 77, 133
cat 203
CD-ROM 89
CEC 3
CEC backplane 21
CEC front view 20
CEC options 101
CEC rear view 21
cell 8
Central Electronics Complex 3, 19
CFReport file 117
CFReport, e-config 117
Chipkill recovery 165
CIM 76
CIU 25
clock card 21
clock signal 21
Cluster 1600 11
cluster authentication daemon 201
CMOS-8S3SOI 24
235
CMOS-9SSOI 24
Collect VPD Information 192
commercial processing 11
Common Information Model 76
concurrent repair 163
Conduct Microcode Survey 191
configuration guidelines 81
configuration rules 81
Connection tab
e-config 106
copper interconnects 4
CPU 8
CPU Guard 170
CPU_FAILURE_PREDICTED 171
CRAD 98
CRC 162
CTA 144
CTA accepted bit 144
ctcas subsystem 201
ctcasd daemon 201
CUoD 9, 133
CUoD activation process 147
CUoD application 150
CUoD capacity card 136
CUoD error codes and messages 146
customer requested arrival date 98
cyclic redundancy checking 162
D
daughter card 47–48
DC 73
DCA 20
DDR 5, 34
Dead Gateway Detection 208
default configuration 83
detailed diagram
e-config 104
detailed message text 115
devices.chrp.base.ServiceRM fileset 198
DGD 208
disk mirroring 174
diskette drive 70
distance solution 76
distributed converter assembly 20
Distributed Management Task Force 76
DLPAR 6, 9, 151
DMA operation 166
DMSRM 202
236
DMTF 76
DNS 203
domains 8
Double Data Rate 5
Dual Data Rate 34
dual-loop mode 57
duplicate 120
duplicate IP address 203
DVD-RAM 89
DVD-RAM drive 76
DVD-ROM 89
dynamic logical partitioning 6, 9, 136, 151
Dynamic Processor Sparing 9
E
e-config 79, 97
EEH 167
eight execution unit 24
Electronic Service Agent 195
electrostatic discharge 49
Enterprise Resource Planning 11
ERP 11
F
Fabric Controller 25
failing processor 154
fault isolation register 161
fault isolation register bit 161
fault recovery 158
FC 2634 72
FC 2737 92
FC 2848 92
FC 2943 94
FC 2944 94
FC 4139 81
FC 4253 72, 89
FC 5251 81, 85
FC 5257 85
FC 6070 92
FC 6071 92
FC 6074 93
FC 6075 92
FC 6186 93
FC 6187 93
FC 6200 93
FC 6201 93
FC 6404 48
FC 6410 48
IBM ^ pSeries 670 and pSeries 690 System Handbook
FC 6418 48
FC 6419 48
FC 7315 94
FC 8137 94
FC 8692 89
FC 8800 92
FC 8841 92
FC2751 55
FC6206 55
FC8396 55
feature conversion 96
FFDC 26
Fibre Channel 54
Field Replaceable Unit 172
FIR 161
First Failure Data Capture 26, 161
floating-point execution unit 24
floating-point operation 24
FLOP 24
FQDN 203
FRU 172
full system partition 92
fully qualified domain name 203
G
graphical user interface on the HMC 76
graphics console 92
grep 203
GX bus slot 21
H
ha_star 171
HAEH 168
half I/O drawer 11, 51
hardware architecture 17
hardware error checker 161
hardware initialization 159
High Availability Event Handler 168
high performance computing 11, 28
high-performance e-business infrastructures 11
HMC 10
HMC attachment, 76
HMC configuration rules 93
HMC1 76
HMC2 76
host 204
Host/IBMLink 97
hostname 203
Hot-Plug LED 175
Hot-Plug PCI adapters 174
Hot-swappable FRUs 173
HPC 11, 28
HPC feature 23
I
I/O 8
I/O book 21
I/O drawer configuration rules 88, 91
I/O drawer options 106
I/O drawer physical placement order 65
I/O drawer subsystem 51
I/O planar 52
IBF 74
IBM Configurator for e-business 79
IBM Hardware Management Console for pSeries
10, 75
IBM Portable Configurator for RS/6000 98
IBM RS/6000 Model 580 27
IBM sales manuals 98
IBM software products 12, 233
IBM.ServiceRM subsystem 198
IDE DVD-ROM drive 72, 89
IDE media to LVD SCSI interface bridge card 72,
89
inappropriate network configurations 203
inconsistent name resolution 203
initial order 97–98
internal battery feature 74
Internal level 1 (L1) cache 4
Inventory Scout daemon 194
Inventory Scout Services 179, 191
invscout 192
invscoutd 194
inward facing 34
inward facing memory card 86
iSeries 7
J
Java 76
Joint Test Action Group 25
JTAG 25
Jump to Wizard 115
K
keyboard/mouse attachment card–PCI 92
Index
237
L
L1 cache 164
L1 data cache 24
L1 instruction cache 24
L2 cache 3, 25, 164
L3 cache 5, 164
L3 cache deconfiguration 172
L3 controller 25
L3 directory 25
LDAP 203
Light Path Diagnostics 176
Linux 3, 13
Linux on pSeries 234
Linux operating system 76
logical partitioning 3, 7–8
long distance solution 94
LPAR 6
LPAR considerations 84
LparCmdRM 202
lsattr 171
lslpp 198
lssrc 198
M
MA 201
machine crash 206
Machine Reported Product Data 150
mainframe system 7
maintenance agreement 201
managed system 75
MCM 3
MCMs and GX slots relationship for pSeries 670 44
MCMs and GX slots relationship for pSeries 690 39
MDA 75
media drawer 70, 83, 89
memory 8
memory cards 34
memory cards configuration for pSeries 670 43
memory cards configuration for pSeries 690 34
memory configuration rules 86
memory controller chip 33
MES 96–97
microcode survey upload file 193
microcode updates 192
minimum configuration 81, 83
mission-critical environments 11
model conversion 96
modular design 17
238
Motor Drive Assembly 75
Motor/Scroll Assembly 75
mouse–black with keyboard attachment cable 92
MRPD 150
MSA 75
multichip module 3, 22, 26
multi-path routing 208
multiple MCM configuration 31
N
name resolution 203
NC 25
network interface backup 208
NIS 203
node name 203
NON BOOT error 198
non-cacheable 25
nPartitions 8
NVRAM 50, 160
O
ODM 171
OLTP 11
Online Transaction Processing 11
OpenSSH 204
operator panel 70
out of compliance 155
out-of-order 24
outward facing 34
outward facing memory card 86
P
parity checking 162
partition page table 86
partition profile 88
partitioning-capable pSeries server 75
PCI 2.0 Specification 167
PCI adapter 54
PCI bus deallocation 166
PCI bus parity error recovery 166
PCI compliance 167
PCI Extended Error Handling 167
PCI slot 51
PCI-X 2
PCRS6000 98
performing an upgrade 116
PERMANENT error 198
IBM ^ pSeries 670 and pSeries 690 System Handbook
physical damage 49
physical location code 67
physical partition 8
pipeline structure 24
POR 26
possible CUoD combinations 139–140
POST 162
power and RAS infrastructure 56
power controller feature 93
POWER GXT135P graphics accelerator 92
Power On Reset 26
POWER2 27
POWER4 chip 3, 22
POWER4 core 24
POWER4+ 2, 22
power-on self test 162
predictive functions 158
price quotation 79
primary battery backup feature 93
primary I/O book 47
problem determination 178–179
processor bus pass through module 29, 85, 105
processor configuration rules 85
Processor Hot Sparing 9, 135
programmable processor clock card 85
Proposed tab 116
ps 198
pSeries 610 7
pSeries 630 8
pSeries 655 7
pSeries 660 Model 6M1 51
pSeries 670 1, 17
pSeries 670 (16-way 1.1 GHz) 117
pSeries 690 1, 17
Copper interconnects 4
pSeries 690 (24-way 1.3 GHz) 119
Q
QBB 8
Quad Building Block 8
quiet touch keyboard–USB 92
R
rack indicator light 199
rack options 111
e-config 111
rack-based server 18
RAS 10, 157
light path diagnostics 176
RAS features 158
Redbooks Web site 234
Contact us xvii
redundancy in components 158
redundant battery backup feature 93
redundant bit steering 165
register rename pool 24
reliability 10, 157
Remote Asynchronous Node 94
removable media bay 70
resource monitoring and control 198
reverse and regular name resolutions 205
RIO connector 52
RIO drawer 52, 54
RIO Loop 57
RIO port 40
RIO Riser card 52
RIO-2 books 48
RIO-2 connector 52
RIO-2 drawer 52, 54
RIO-2 Loop 57
riser 47–48
RMC 198
RMC framework 201
rmcctrl 205
rmrsrc 205
rootvg mirroring 207
rsct.core.sec 201
S
S/370 7
SBE 172
scrubbing 165
SCSI 4-pack backplane 51
SCSI Enclosure Services 51
sculptured black front door with copper accent 92
SDRAM 34
secondary I/O books 47
server consolidation 11
Service Agent 179, 195
Service Agent Call-Home capability 77
Service Agent gateway process 196
Service Focal Point 77, 179, 198
service processor 50, 159
serviceability 10, 157
serviceability features 158
SES 51
Index
239
shared L2 cache 4
shconf 207
shdaemon 207
Silicon-on-Insulator 4
single bit error 172
single MCM configuration 32
single points of failure 163
single-loop mode 57
slim line rack door 93
Small Real Mode Address Region 88
SMI 33
SMIT 170
SMP system 8
Software Global Options 99
Software Products 99
SOI 4
SOI CMOS technology 24
SP 11
SP Cluster 1600 considerations 94
SSA 54
standby processor 154
Storage tab 103
e-config 106
superscalar 24
SuSE 13
synchronous DRAM 34
synchronous memory interface 33
Synchronous Wave Pipeline Interface 34
system board 8
system hang detection 207
system IPL 159
system resources 3
un-resolvable host name 203
upgrade considerations 96
Upgrade/MES 116
UPS 163
V
validating the configuration 114
Virtual Terminal 77
vital product data 192
VM 7
VPD 192
W
work area 99
Z
zSeries 3
zSeries LPAR implementation 8
T
TCE 86
TCP/IP RAS enhancements 208
third-party software products 12, 233
transistor 4
translation control entry 86
trouble-free network planning rules 203
two-way activation 136
U
Ultra3 SCSI disk drive 53
uninterruptible power supply 163
UNIX hostname authentication 203
unreachable network interface 204
un-resolvable 203
240
IBM ^ pSeries 670 and pSeries 690 System Handbook
IBM ^ pSeries 670 and pSeries 690 System Handbook
(0.5” spine)
0.475”<->0.875”
250 <-> 459 pages
Back cover
®
pSeries 670
IBM
and pSeries 690
System Handbook
Component-based
description of the
hardware
architecture
How to use e-config
to configure pSeries
670 and pSeries 690
Capacity Upgrade on
Demand explained
This IBM Redbook explains the IBM ^ pSeries 670
Model 671 and the IBM ^ pSeries 690 , a new level
of UNIX servers providing world-class performance,
availability, and flexibility. Capable of data center, application,
and high performance computing, these servers include
mainframe-inspired self-management and security to meet
your most demanding needs.
This redbook includes the following topics:
-Overview of the pSeries 670 and pSeries 690
-Hardware architecture of the pSeries 670 and pSeries 690
-Using the IBM Configurator for e-business
-Capacity Upgrade on Demand
-Reliability, availability, and serviceability
This book is an ideal desk-side reference for IBM
professionals and Business Partners, and technical
specialists who support the pSeries 670 and pSeries 690, and
for those who want to learn more about this radically new
server in a clear, single-source handbook. It provides the
necessary information to successfully order a pSeries 670 or
pSeries 690 for a production environment and then, upon
successful installation, configure service support functions,
such as Service Focal Point and Inventory Scout.
INTERNATIONAL
TECHNICAL
SUPPORT
ORGANIZATION
BUILDING TECHNICAL
INFORMATION BASED ON
PRACTICAL EXPERIENCE
IBM Redbooks are developed by
the IBM International Technical
Support Organization. Experts
from IBM, Customers and
Partners from around the world
create timely technical
information based on realistic
scenarios. Specific
recommendations are provided
to help you implement IT
solutions more effectively in
your environment.
For more information:
ibm.com/redbooks
SG24-7040-02
ISBN 0738499501
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement