Dell PowerEdge 12th Generation Server System

Dell PowerEdge 12th Generation Server System
Dell PowerEdge 12th Generation Server
System Profiles
A Dell technical white paper that shows the performance and energy
efficiency differences between the user-selectable BIOS System Profiles
Bruce Wagner
Solutions Performance Analysis
Dell PowerEdge 12th Generation Server System Profile Study
Contents
Executive summary ..................................................................................................... 3
Introduction .............................................................................................................. 3
Test Methodology ....................................................................................................... 7
Results 1: Performance per Watt (DAPC) versus Performance ............................................. 8
Results 2: Performance per Watt (DAPC) versus Performance per Watt (OS)........................... 10
Conclusions .............................................................................................................. 12
Appendix A: System Under Test (SUT) configuration .......................................................... 13
Appendix B: Power and Cooling Calculation Assumptions .................................................... 14
Appendix C: References ............................................................................................ 15
Tables
Table 1.
“Easy Button” BIOS settings correspondence ........................................................ 3
Table 2.
Compute throughput versus input power across the full workload range ....................... 8
Table 3.
Compute throughput versus input power across the full workload range ...................... 10
Table 4.
Detailed system test configuration ................................................................... 13
Figures
Figure 1.
POST splash screen with F2 = System Setup ......................................................... 4
Figure 2.
BIOS Settings menu ...................................................................................... 4
Figure 3.
System Profile Settings menu .......................................................................... 5
Figure 4.
Microsoft Windows Server 2008 power plan menus ................................................. 6
Figure 5.
Microsoft Windows Server 2008 Power Saver plan settings........................................ 6
Figure 6.
SPECpower test bed ..................................................................................... 7
Figure 7.
Compute throughput versus system power consumption .......................................... 9
Figure 8.
Compute throughput versus system power consumption ......................................... 11
Figure 9.
Annual operating cost of servers per watt .......................................................... 14
This document is for informational purposes only and may contain typographical errors and technical
inaccuracies. The content is provided as is, without express or implied warranties of any kind.
© 2014 Dell Inc. All rights reserved. Dell and its affiliates cannot be responsible for errors or omissions in
typography or photography. Dell, the Dell logo, and PowerEdge are trademarks of Dell Inc. Intel and Xeon
are registered trademarks of Intel Corporation in the U.S. and other countries. Microsoft, Windows, and
Windows Server are either trademarks or registered trademarks of Microsoft Corporation in the United
States and/or other countries. Other trademarks and trade names may be used in this document to refer
to either the entities claiming the marks and names or their products. Dell disclaims proprietary interest
in the marks and names of others.
January 2014 | Rev 1.10
ii
Dell PowerEdge 12th Generation Server System Profile Study
Executive summary
Customers choosing one of the performance-per-watt optimized system profiles instead of the performance-optimized
one can experience up to 60% reduced system idle power consumption and an overall power efficiency increase of 19%
when the system if fully utilized. Given industry research that an Enterprise server’s annual fully burdened cost per
watt is $2.69, a savings of $134 per server, per year can be realized.
Introduction
The new Dell™ PowerEdge™ 12th generation servers support Dell Active Power Controller (DAPC) system profiles that
serve as “easy buttons” for applying default BIOS and firmware values targeted for performance, performance per
watt, or RAS for dense configurations. Table 1 shows the BIOS settings for each System Profile.
Table 1.
“Easy Button” BIOS settings correspondence
System Profile
Subsystem
CPU
Memory
Thermal
Setting
Performance
Performance
per watt (OS)
Performance
per watt
(DAPC) (default)
Dense
configuration
Custom
CPU Power
Management
Maximum
Performance
OS DBPM
System DBPM (DAPC)
System DBPM
(DAPC)
System
DBPM
(DAPC)
Turbo Boost
Enabled
Enabled
Enabled
Disabled
Enabled
C1E
Disabled
Enabled
Enabled
Enabled
Enabled
C-States
Disabled
Enabled
Enabled
Enabled
Enabled
Frequency
Maximum
Performance
Maximum
Performance
Maximum
Performance
Dense
Configuration
Optimized
Maximum
Performance
Voltage
Auto
Auto
Auto
Maximum
Auto
Patrol Scrub
Standard
Standard
Standard
Extended
Standard
BIST
Duration
Standard
Standard
Standard
Extended
Standard
Refresh
Rate
1x
1x
1x
2x
1x
Thermal
Algorithm
Maximum
Performance
Minimum
Power
Minimum Power
Dense
Configuration
Maximum
Performance
3
Dell PowerEdge 12th Generation Server System Profile Study
The system profiles can be selected through the following BIOS menus in Figure 1, Figure 2, and Figure 3.
Figure 1.
POST splash screen with F2 = System Setup
Figure 2.
BIOS Settings menu
4
Dell PowerEdge 12th Generation Server System Profile Study
Figure 3.
System Profile Settings menu
5
Dell PowerEdge 12th Generation Server System Profile Study
®
®
Microsoft Windows Server power plans can be adjusted through Control Panel menus, shown in Figure 4 and Figure 5:
Figure 4.
Figure 5.
Microsoft Windows Server 2008 power plan menus
Microsoft Windows Server 2008 Power Saver plan settings
6
Dell PowerEdge 12th Generation Server System Profile Study
Test Methodology
We used the SPECpower_ssj2008 industry-standard benchmark1 to compare server power efficiency results for each of
the “easy button” user-selectable system profiles.
The SPECpower_ssj2008 metric is a calculation of the performance-to-power ratio across the full range of system
workload levels from active idle to 100% utilization. The workload itself models transactions commonly associated with
managing a business warehouse including new order, payment, order status, delivery, stock level, and customer report
generation.
Each measurement point is of the number of ssj_ops (server side Java operations per second) processed per watt of
power consumed. The more ssj_ops the system under test (SUT) can output for a given number of watts of power
input; the better its efficiency.
The SPECpower measurement test bed shown in Figure 6 consists of four hardware components working together to
collect a server's power consumption and performance data while the SUT is exercised with a workload exercising the
SUT with a predefined workload :

Server Under Test (SUT) is the system driven by the SSJ-based workload. The SUT’s performance and power
consumption characteristics are captured and measured by the benchmark.

Power Analyzer is used to measure and record the power consumed by the SUT.

Temperature Sensor is used to capture the temperature of the environment where the SUT is being
benchmarked.

Controller System is a separate system that applies the SUT workload while capturing power and temperature
measurements.
Figure 6.
SPECpower test bed
1
See Appendix C: References for a link to the Standard Performance Evaluation Corporation’s SPECpower_ssj2008
website for more information about the SPECpower_ssj2008 benchmark.
7
Dell PowerEdge 12th Generation Server System Profile Study
Results 1: Performance per Watt (DAPC) versus Performance
Table 2.
Compute throughput versus input power across the full workload range
Target load
(Perf/Watt
DAPC) ssj_ops
(Perf/Watt
DAPC) Power
(Performance)
ssj_ops
(Performance)
Power
100%
866,247
212
875,480
214
90%
787,737
194
791,399
207
80%
703,517
182
705,736
199
70%
613,679
167
618,043
192
60%
524,531
152
528,858
183
50%
437,506
140
441,469
176
40%
350,215
132
352,899
168
30%
262,010
121
264,335
160
20%
174,960
114
175,169
151
10%
86,747
105
88,520
143
Active Idle
0
∑ssj_ops / ∑power =
83.6
3,000
0
134
2,513
8
Dell PowerEdge 12th Generation Server System Profile Study
Figure 7.
Compute throughput versus system power consumption
Summarizing the system behavior differences between Performance per Watt (Dell Active Power Controller) and
Performance-optimized profiles:

Performance per watt (DAPC) system profile idle workload power advantage over Performance (W)
= 50.4

Performance per watt (DAPC) system profile idle workload power advantage over Performance (%)
= 60%

Performance per watt (DAPC) system profile overall efficiency advantage over Performance (%)
= 19%
9
Dell PowerEdge 12th Generation Server System Profile Study
Results 2: Performance per Watt (DAPC) versus Performance per Watt (OS)
Table 3.
Compute throughput versus input power across the full workload range
Target
Load
(Perf/Watt
DAPC)
ssj_ops
(Perf/Watt
DAPC)
Power
(Perf/Watt
OS-power
saver
default
5/75%)
ssj_ops
100%
866,247
212
589,965
153
875241
215
866,247
215
879,249
215
90%
787,737
194
534,918
149
795212
203
787,737
206
795,177
207
80%
703,517
182
476,384
144
706992
182
703,517
185
704,295
199
70%
613,679
167
416,501
138
614191
172
613,679
174
620,136
191
60%
524,531
152
354,536
131
528268
160
524,531
164
529,418
182
50%
437,506
140
296,630
125
441732
145
437,506
152
443,465
173
40%
350,215
132
236,329
118
349904
132
350,215
139
354,411
163
30%
262,010
121
177,567
113
264109
121
262,010
127
266,751
151
20%
174,960
114
119,302
108
176166
113
174,960
116
176,732
137
10%
86,747
105
59,788
101
87413
105
86,747
106
88,166
118
Active
Idle
0
83.6
0
83.2
0
83
0
83.6
0
84.7
∑ssj_ops / ∑power =
3,000
(Perf/Watt
OS-power
saver
default
5/75%)
Power
(Perf/Watt
OS-power
saver
optimum
0/100%)
ssj_ops
(Perf/Watt
OS- power
saver
optimum
0/100%)
Power
(Perf/Watt
OSbalanced)
ssj_ops
2,393
2,966
(Perf/Watt
(Perf/Watt (Perf/Watt
OSOS-highperf)
OS-high
balanced)
ssj_ops
perf) Power
Power
2,899
2,668
10
Dell PowerEdge 12th Generation Server System Profile Study
Figure 8.
Compute throughput versus system power consumption
Key findings from a comparison of system responses under Performance Per Watt (DAPC) and Performance Per Watt
(OS Windows Server 2008 R2 Enterprise) profile:

Performance per Watt (DAPC) vs Performance per Watt (OS) – power saver 0/100% plan overall efficiency advantage (%)
= 1%

Performance per Watt (DAPC) vs Performance per Watt (OS) – balanced plan overall efficiency advantage (%)
= 3%

Performance per Watt (DAPC) vs Performance per Watt (OS) – high performance plan overall efficiency advantage (%)
= 12%

Performance per Watt (DAPC) vs Performance per Watt (OS) – power saver default 5/75% plan overall efficiency advantage (%)
= 25%
11
Dell PowerEdge 12th Generation Server System Profile Study
Conclusions
Choosing the performance-per-watt optimized system profile instead of performance-optimized option reduces system
idle power consumption by 60% while improving overall energy efficiency by 19% and impacting only peak compute
performance by 1%.
The Dell Active Power Controller (DAPC) profile built into every PowerEdge server achieves even better overall energy
efficiency than an OS-based one such as in Windows Server 2008 R2 Enterprise.
Version 1.10 Addendum
It is important to note that the performance-per-watt power management control algorithms evaluated here, base
their adaptive system resource throttling decisions upon the CPU’s reported total utilization at any given moment.
Light workloads (e.g. ones that recruit fewer process threads than available CPU cores) may not exhibit sufficient total
CPU utilization to trigger an instantaneous return to the system’s highest throughput/lowest latency levels. For more
detail, see the “12G PowerEdge Power Efficiency ‘How To’ v1.0” white paper listed in the appendix.
The SPECpower_ssj2008 metric used in this study is fully threaded and thereby models a server at its maximum
potential just as it would be in a typical consolidated datacenter environment. Lightly-threaded, latency-sensitive
application performance with less regard for system power consumption, may be better served by selecting alternate
system profile settings.
12
Dell PowerEdge 12th Generation Server System Profile Study
Appendix A: System Under Test (SUT) configuration
The server configuration chosen for this study is one that was considered “typical” in that it closely resembles what
customers were currently specifying for datacenter deployment. We used this same configuration for Dell’s 12th
generation performance and power claims, and total cost of ownership (TCO) studies.
Table 4.
Detailed system test configuration
Platform
Dell PowerEdge R620
Processor model
2 x Intel® Xeon® E5-2620
Processor frequency
2000 MHz
Processor L3 cache
15 MB
Physical cores
12
Logical processors
24
Memory frequency
1333 MHz
Memory details
8 x 4 GB 2Rx4 LV RDIMMs
Internal Storage
2 x 146GB 15K 2.5in SAS HDD
HBA
PERC H710p mini
BIOS version
1.2.1
iDRAC version
iDRAC7 Enterprise 1.06.06 (Build 15)
NDC
4 x 1Gb Base-T
Power supply quantity and rating
2 x 495W
Line Voltage
208VAC
Turbo Boost
Enabled
Processor prefetchers
Disabled
Logical processors
Enabled
Node interleaving
Disabled
C-states
Set by system profile
C1E
Set by system profile
Power management
Set by system profile
13
Dell PowerEdge 12th Generation Server System Profile Study
Appendix B: Power and Cooling Calculation Assumptions
To calculate a server’s annual fully burdened cost per watt, we used an industry expert’s formula2 updated with a
$0.11/kWh cost of power to come up with the figure of $2.69 per watt per year.
Figure 9.
2
Annual operating cost of servers per watt
See Appendix C: References for a link to the article “Annual Fully Burdened Cost of Power” by James Hamilton.
14
Dell PowerEdge 12th Generation Server System Profile Study
Appendix C: References
-
The Standard Performance Evaluation Corporation’s SPECpower_ssj2008 metric for volume, multi-node class
computer efficiency. http://www.spec.org/power_ssj2008/
-
Dell white paper: “12G PowerEdge Power Efficiency ‘How To’ v1.0” by John Jenne.
http://en.community.dell.com/techcenter/extras/m/white_papers/20109038.aspx
-
Dell white paper: “Dell Energy Smart Architecture and Power Management Adoption” by Brad Lawrence, John
Jenne. http://en.community.dell.com/techcenter/extras/m/white_papers/20109094.aspx
-
Dell white paper: “Right-Sized Power Systems: A Means to Improved Energy Efficiency” by Eric Wilcox, Mark
Muccini http://en.community.dell.com/techcenter/extras/m/white_papers/20106404.aspx
-
U.S. EPA “Energy Star Program Requirements for Computer Servers”, v2.0. http://www.energystar.gov
-
80 PLUS computer power supply initiative. http://www.plugloadsolutions.com/80PlusPowerSupplies.aspx
-
Annual Fully Burdened Cost of Power; James Hamilton; 2008:
http://perspectives.mvdirona.com/2008/12/06/AnnualFullyBurdenedCostOfPower.aspx
-
Dell Server BIOS LLD specification: “12G Dell Active Power Controller (DAPC) v1.1” by Wuxian Wu
-
Dell’s Power and Cooling Technologies website
-
Microsoft OS Power Plan primer: http://windows.microsoft.com/en-us/windows7/Power-plans-frequentlyasked-questions
-
Dell PowerEdge R620 12th Generation Rack Server: http://www.dell.com/us/business/p/poweredge-r620/pd
15
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

advertising