Sun Fire T1000 and T2000 Server Architecture

SUN FIRE T1000 and T2000
SERVER ARCHITECTURE
Unleashing the UltraSPARC T1 Processor with
CoolThreads Technology
White Paper
December 2005
Table of Contents
Sun Microsystems, Inc.
Table of Contents
Executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Throughput Computing Realized . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Business Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2
The Diminishing Returns of Traditional Processor Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3
Throughput Computing from Sun: A new Approach to Scalability and Eco-Responsibility . . . . . . . . . . . . . . . . . . . . . . .4
Sun Fire‰ Servers Based on the UltraSPARC T1 Processor with CoolThreads‰ Technology . . . . . . . . . . . . . . . . . . . . .5
Space, Watts, and Power: Introducing the SWaP Metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7
The UltraSPARC T1 Processor with CoolThreads Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Chip Multithreaded (CMT) Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9
UltraSPARC T1 Processor Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Unsurpassed Reliability, Availability, and Serviceability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Sun Fire T1000 and T2000 Server Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Sun Fire T1000 Server Architecture Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Sun Fire T2000 Server Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Getting the Most from Throughput Computing:
A Complete and Cohesive Software and Management Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
The Solaris‰ Operating System: Scalability and Support for CoolThreads Technology . . . . . . . . . . . . . . . . . . . . . . . . 24
The Java Enterprise System (JES) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Sun Studio 11 Compilers and Analysis Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
System Management . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1-Executive Summary
Sun Microsystems, Inc.
Executive Summary
Marked by the prevalence of web services and service-oriented architecture (SOA), the emerging Participation Age
promises the ability to deliver rich new content and high-bandwidth services to larger numbers of users than ever
before. Through this transition, organizations across many industries hope to address larger markets, reduce costs,
and gain better insights into their customers. At the same time, an increasingly broad array of wired and wireless
client devices are bringing network computing into the everyday lives of millions of people, redefining data center
scalability and capacity requirements.
While these developments offer considerable potential benefits, many organizations face the seemingly
contradictory challenges of scaling their services and improving utilization within a fixed envelope of space, power,
and cooling. Even as they struggle to add performance and capacity, real hard limits of available real estate and
power are abundantly clear. In fact, IDC now predicts a 50 percent increase in the installed base of servers in the
United States alone by 2009, further estimating the annual power bill for that installed base at five billion dollars1
(excluding cooling). Gartner now estimates that more than 80 percent of all data centers are already constrained
by electrical power, physical space, or cooling capacity2. Simply adding more hot and power-hungry systems is
clearly no solution, only contributing to new levels of consumption, complexity, and waste. Efficient power and
cooling are critical, indicated by data center managers who recently reported a 10 percent or more increase in
power requirements over the past year, with fully 41 percent reporting that they anticipated upgrading their power
and cooling systems over the next three years.3
Consistent with Sun's Throughput Computing initiative and eco-responsibility commitment, Sun Fire T1000 and
T2000 servers with CoolThreads technology provide a welcome new approach. These systems are powered by the
breakthrough innovation provided by the UltraSPARC T1 processor and they completely change the equation on
space, power, and cooling in the data center. The UltraSPARC T1 processor delivers up to seven times the
performance of its competitors4, effectively condensing the resources of a large symmetric multiprocessing (SMP)
system onto a single chip. Supporting up to 32 active threads, the UltraSPARC T1 processor offers massive
improvements in application scalability and processor utilization—presenting opportunities for considerable
application consolidation while requiring only about as much energy to operate as a common light bulb.
Best of all, Sun is bringing this cutting-edge technology to market in the familiar and proven SPARC architecture
and the compact and reliable enterprise-grade Sun Fire T1000 and T2000 servers. Thanks to Sun’s Binary
Compatibility Guarantee and the robust Solaris Operating System (OS), the increased resources and increased
utilization from these systems are immediately available to existing applications and environments. With its strong
research and development investments and extensive intellectual property in processor and system design,
operating systems, middleware and management software, Sun is uniquely positioned to rapidly bring the benefits
of Throughput Computing to its customers, without requiring so much as a recompile.
1. "Server Power Consumption Reemerges as a Critical Cost Factor in Datacenters", Vernon Turner, IDC, August 2005
2. Informal poll from Gartner’s 2004 Datacenter Conference, as reported in "Dealing with a Less Than Perfect Data Center Location",
by Johanna Ambrosio, Data Center Futures, TechTarget mailing, January 21, 2005
3. Joint study published by AFCOM and InterUnity Group in April 2005, and reported on at
http://searchdatacenter.techtarget.com/originalContent/0,289142,sid80_gci1081270,00.html
4.RSA and DSA sign operation @1024-bit, based on cryptography performance tests
2-Throughput Computing Realized
Sun Microsystems, Inc.
Chapter 1
Throughput Computing Realized
Businesses today are increasingly defined by their applications, and now more than ever, an organization’s
prospects for success are increasingly fixed to its ability to deploy technology in an agile and effective fashion. The
risks are extreme. In today’s competitive and highly-regulated business environment, the cost of technology failure
can be rapid and severe. Even small lapses in IT competence can result in wide-spread damage and loss.
Business Requirements
Increasing the pressure, an endless variety of new networked devices and users are demanding ever-higher levels
of performance, capacity, availability, and security from the applications and services that serve them. Real estate
concerns along with very real and rising energy costs for both power and cooling are now significant factors that
discourage merely adding endless racks of traditional servers. The cost and complexity of managing very large
numbers of systems is another pressing concern, especially when coupled with the very low levels of utilization
typically found in traditional infrastructure.
To respond to these myriad challenges, business must:
• Grow infrastructure and increase application throughput, capacity, and performance to address pressing
business needs as well as capture new customers and opportunities
• Reduce power, cooling, and real estate costs both to save money and to facilitate necessary growth and
scalability for the future
• Push for effective consolidation to counter high complexity and management costs, resulting in fewer systems
that accomplish more work
• Increase virtualization of resources to aid with business agility and improved resource utilization
• Maintain application compatibility and enhance security across the organization to preserve investments and
limit risks to the firm and its clientele
Beyond mere packaging, these issues drive to the very technology used to design processors, systems, and
applications. Processor design in particular can have enormous ramifications for business-level issues and
solutions. Unfortunately, traditional high-frequency, single-threaded processors are increasingly yielding
diminishing returns. Even with ever-higher clock rates, these processors are producing only small improvements in
real-world application performance. At the same time, these high-frequency processors generate escalating costs
in the form of higher levels of power consumption, and significantly higher levels of heat load that must be
addressed by multiple large and expensive HVAC systems. With economic and competitive pressures at an all-time
high, most understand that significant change is needed.
While optimistic marketing statements constantly call attention to presumably impressive multiple-gigahertz
frequencies and high levels of cache for new generations of processors, corresponding small gains in real-world
system performance and productivity continue to frustrate IT professionals. Throughput Computing, along with
Sun’s focus on optimizing real workload performance is designed to help resolve these divergent trends. This
approach provides higher levels of delivered performance and computational throughput while greatly simplifying
3-Throughput Computing Realized
Sun Microsystems, Inc.
the data center. Understanding the importance of throughput computing requires a look at how both processors
and systems have been designed in the past, and the trends that are defining better ways forward.
The Diminishing Returns of Traditional Processor Design
The oft-quoted tenant of Moore’s Law states that the number of transistors that will fit in a square inch of
integrated circuitry will approximately double every two years. For over three decades the pace of Moore’s law has
held, driving processor performance to new heights. Processor manufacturers have long exploited these chip real
estate gains to build increasingly complex processors, with instruction-level parallelism (ILP) as a goal. Today these
traditional processors employ very high frequencies along with a variety of sophisticated tactics to accelerate a
single instruction pipeline, including:
•
•
•
•
•
•
Large caches
Superscalar designs
Out-of-order execution
Very high clock rates
Deep pipelines
Speculative pre-fetches
While these techniques have produced faster processors with impressive-sounding multiple-gigahertz frequencies,
they have largely resulted in complex, hot, and power-hungry processors that don’t serve many modern
applications, or the constraints of today’s data centers. In fact, many of today’s data center workloads are simply
unable to take advantage of the hard-won ILP provided in these processors. As shown in Table 1, applications with
high shared memory and data requirements are typically more focused on processing a large number of
simultaneous threads (thread-level parallelism, TLP) rather than running a single thread as quickly as possible
(ILP).
Table 1. Attributes of common commercial workloads favor thread-level parallelism over instruction-level
parallelism
Workload Attributes
Web-Centric
Application-Centric
Data-Centric
Benchmark
Web
(SPECweb99)
Application
SAP-SD
(SPECjAppServer2002) 2Tier
Data
(TPC-C)
SAP-SD
3Tier (DB)
DSS
(TPC-H)
Application category
Web server
Server, Java
ERP
OLTP
ERP
DSS
Instruction-level parallelism
Low
Low
Medium
Low
Low
High
Thread-level parallelism
High
High
High
High
High
High
Instruction/Data working set
Large
Large
Medium
Large
Large
Large
Data sharing
Low
Medium
Medium
High
High
Medium
(SPEC, SPECweb, SPECjAppServer, and SPECjbb are registered trademarks of the Standard Performance Evaluation Corporation (SPEC).
SAP is a registered trademark of SAP AG in Germany and other countries. TPC-C and TPC-H are trademarks of the Transaction
Processing Performance Council (TPC). For more information see www.spec.org, www.sap.com/benchmark, and www.tpc.org.)
Complicating matters, the disparity between processor speeds and memory access speeds means that memory
latency dominates application performance, erasing even very impressive gains in clock rates. While processor
speeds continue to double every two years, memory speeds have typically doubled only every six years. This
growing disconnect is the result of memory suppliers focusing on density and cost as their design center, rather
4-Throughput Computing Realized
Sun Microsystems, Inc.
than speed. Unfortunately, this relative gap between processor and memory speeds leaves ultra-fast processors
idle as much as 85 percent of the time, waiting for memory to return required data. Ironically, as traditional
processor execution pipelines get faster and more complex, the effect of memory latency grows—fast, expensive
processors spend more cycles doing nothing. Worse still, idle processors continue to draw power and generate
heat. Its easy to see that frequency (gigahertz) is truly a misleading indicator of real performance.
Figure 1 illustrates how even doubling processor performance (frequency) often provides only a small relative
increase in application performance. In this example, though the compute time is reduced by half, only a small
overall improvement in execution time results, due to the constant and dominant influence of memory latency.
Thread
C
M
C
M
C
C
Compute time
M
Memory latency
M
Time
Thread
C
M
C
M
C
M
Time
Time saved with a 100 percent
increase in processor performance
Figure 1. Increasing single-threaded processor performance by 100 percent (a 50-percent reduction in
compute-time) provides only a small relative gain in application performance due to memory latency
While some vendors have seemingly awakened to the inherent limitations of traditional, frequency-based
processor designs, they are now attempting to graft power-saving technologies and multiple cores onto old, oncediscarded architectures. Unfortunately, these efforts represent stop-gap measures at best. Effective approaches
can only be realized with fundamentally new processor designs that deliver truly compelling benefits.
Throughput Computing from Sun: A new Approach to Scalability and Eco-Responsibility
In the face of these challenges, Sun's Throughput Computing strategy provides an effective combination of
processor, platform, OS, and application technology. The result is a systemic focus on the performance of key
workloads, not just the frequency of the processor, or the performance of a single thread of execution:
• With up to seven times the performance of other modern processors, UltraSPARC T1 processors with
CoolThreads technology provide massive amounts of thread level parallelism and increased application
throughput through chip multithreading (CMT).
• Sun Fire T1000 and T2000 servers deliver the considerable resources of UltraSPARC T1 processors in reliable
enterprise-class 1U and 2U platforms while significantly improving power consumption and heat generation.
• The innovative and proven Solaris OS facilitates consolidation through fine-granularity partitioning and
virtualization, security, and very high levels of utilization while staunchly upholding Sun's binary compatibility
guarantee.
• Compilers, development tools, middleware, and an end-to-end systems strategies leverage the resources of
innovative CMT processors, resulting in real-world application performance improvements.
5-Throughput Computing Realized
Sun Microsystems, Inc.
Figure 2. The UltraSPARC T1 Processor with CoolThreads technology
Sun Fire Servers Based on the UltraSPARC T1 Processor with CoolThreads Technology
Sun Fire T1000 and T2000 servers represent clean-sheet designs that leverage the considerable resources of the
UltraSPARC T1 processor (Figure 2) to address the challenges of modern data centers. Optimized for network
workloads, these systems deliver up to three times the throughput of competitive systems, while yielding up to
four times better performance per watt1, and up to four times better SWaP (as evaluated by the Space, Watts, and
Power metric described later in this document).
With large memory support and the breakthrough performance of the UltraSPARC T1 processor, these systems are
ideally suited for compute, data, and transaction-intensive applications. With a six- or eight-core UltraSPARC T1
processor and dense 1U packaging, the Sun Fire T1000 server (Figure 3) is an ideal platform for web, portal,
network or security servers. With considerable expansion capabilities and features such as hot-swappable
redundant fans, power supplies, and disk drives, the 2U Sun Fire T2000 server is ideal for OLTP, CRM, ERP, database,
and collaboration. Both servers excel in the delivery of Java based applications and services, as proven by a suite of
world record benchmarks at the product introduction in December 2006 (http://www.sun.com/servers/
sparc_benchmarks/).
Sun Fire T1000 Server
Sun Fire T2000 Server
Figure 3. Sun Fire T1000 and T2000 servers with CoolThreads Technology
1.Disclosure: SPEC and SPECweb are registered trademarks of the Standard Performance Evaluation Corporation. Sun Fire T2000
server (8 cores, 1 chip) 14001 SPECWeb2005. IBM eServer Xseries x346 (2 cores, 2 chips) 4348 SPECWeb2005. Sun Fire T2000 server
results submitted to SPEC and other results from www.spec.org as of 12/06/05. IBM x346 specifications from product brochure,
09/05/05: http://www-03.ibm.com/servers/eserver/xseries/x346.html. IBM x346 power rating estimated by calculating 70% of the
power supply data reported in the product brochure. Sun Fire T2000 server power consumption taken from measurements made
during the benchmark run.
6-Throughput Computing Realized
Sun Microsystems, Inc.
Designed to complement each other, the Sun Fire T1000 and T2000 servers share capabilities that are essential to
addressing the needs of the modern data center:
• Breakthrough performance in less space:
With support for up to 32 compute threads in a single processor, Sun Fire T1000 and T2000 servers provide very
high compute density in a very small footprint, easing the limitations that many data centers are currently
experiencing. In fact, according to Sun’s analysis, companies using UltraSPARC T1 processor-based systems can
experience a 10:1 reduction in the number of servers they already have deployed1.
• Lower power costs with CoolThreads technology:
CoolThreads technology in the UltraSPARC T1 processor requires significantly less power than competitive
processors, making them substantially less expensive to operate and facilitating greater levels of capacity in
existing data center facilities. The UltraSPARC T1 processor draws as little as a typical 72 watts per processor,
compared with competitive processors that require about the same amount of power per thread. This crucial
difference means that Sun Fire T1000 and T200 servers provide up to five times the performance per watt of
competing servers2. Power cost reductions alone from deploying Sun Fire T1000 and T2000 servers can save
organizations millions of dollars annually.
• Increased reliability through drastic simplification:
Consolidation of business-critical applications requires platforms that can deliver very high levels of reliability,
availability, and serviceability (RAS). Because the UltraSPARC T1 processor replaces many processors and
interconnects found in traditional SMP systems, it can deliver very high levels of RAS efficiency. In addition, Sun
Fire T1000 and T2000 servers use far fewer components than competing systems, greatly improving on service
intervals. The UltraSPARC T1 processor coupled with the Solaris OS also offer a wealth of technologies to
improve reliability, availability and serviceability (detailed later in this document).
• Greater application efficiency, security, and investment protection:
The Sun Fire T1000 an T2000 servers come installed with the Solaris 10 OS, providing an efficient and secure
application environment while further increasing performance and utilization. The Solaris 10 OS multithreaded
design takes full advantage of these server’s chip multithreaded (CMT) architecture, boosting throughput and
efficiency. Solaris Containers consist of a group of technologies that work together to efficiently manage the
considerable resources of the UltraSPARC T1 processor, virtualize the environment, and provide a complete,
isolated, and secure runtime environment for multiple applications on a single server. The Solaris 10 OS protects
against both malicious external attacks and data access violations from the inside, and of course UltraSPARC T1
based systems are fully binary compatible with other SPARC based Solaris systems, and source-code compatible
with Solaris based systems running on any architecture.
1.Results achieved with AMT-C tool, a Sun consolidation ROI tool. The comparison was based on 200 older Dell PE 2650 servers with
2x2.4GHz Intel Xeon processors. The consolidation was based on 20 Sun Fire T1000 servers running the Solaris 10 OS and Solaris
Containers. The Dell system average processor utilization was 25% compared to an 80% utilization for the consolidated Sun Fire
T1000 servers.
2.Disclosure: SPEC and the benchmark name SPECjAppServer are trademarks of the Standard Performance Evaluation Corporation.
Comparison between results based on the SPECjAppServer2004 benchmark. The Sun Fire T2000 server (8 cores, 1 chip) achieved
615.64 JOPS@Standard while the HP rx4600 server (4 cores, 4 chips) achived 471.28 JOPS@Standard. Sun Fire T2000 results submitted
to SPEC as of 12/06/05. Other results from www.spec.org as of 12/06/05. HP rx4640 server specifications taken on 10/19/05 from
http://www.hp.com/products1/servers/integrity/engry_level/rx4640/index.html. HP rx4640 power rating of 1,303 watts taken
from HP Enterprise Configurator, 10/19/05. HP rx4640 configured with redundant power, 4x1.6 GHz Itanium processors, 8x2GB
DIMMs, no PCI cards, and 2x73GB HDDs. Sun Fire T2000 server power consumption of 320 watts taken from actual measurements
made during the benchmark run.
7-Throughput Computing Realized
Sun Microsystems, Inc.
Table 2 compares the features of Sun Fire T1000 and T2000 servers.
Table 2. –Comparison of Sun Fire T1000 and T2000 servers
Feature
Sun Fire T1000 Server
Sun Fire T2000 Server
CPUs
Six- or eight-core 1.0 GHz
UltraSPARC T1 processor
Four-, six-, or eight-core 1.0 or 1.2
GHz UltraSPARC T1 processor
Active threads
Up to 32
Up to 32
Maximum memory
16 GB
32 GB
Maximum internal disk drives
One 80 GB SATA disk
Four 73 GB SFF SAS disks
Removable/plugable I/O
N/A
Slimline DVD-R/CD-RW
Four USB 1.1 ports
PCI
One PCI-E slot, low profile
Three PCI-E slots, low profile
Two PCI-X slots, low profilea
Ethernet
Four on-board Gigabit Ethernet ports Four on-board Gigabit Ethernet ports
(10/100/1000)
(10/100/1000)
Power supplies
One 300W power supply
Two redundant (N+1) hot-swap 550W
power suppliesb
Fans
Single fan tray assembly
3 redundant, hot-swappable
cooling fans
Form factor
1 rack unit (1U)
2 rack units (2U)
a. One PCI-X slot occupied with a disk controller
b. 450 watt power supplies will be available in early 2006
Space, Watts, and Power: Introducing the SWaP Metric
Sun Fire T1000 and T2000 servers deliver leading performance across a range of web and application tier
benchmarks. However, with energy and real estate costs and pressures, it is not enough to measure performance
in isolation. Delivering the required level of throughput in a fixed space and power envelope is critical. Traditional
system-to-system benchmarks are valuable as a way of comparing one system to another, but are limited when it
comes to understanding the power and density attributes of the systems being compared.
For this reason, Sun has developed the SWaP metric, standing for Space, Watts, and Performance. Designed to
provide a simple and transparent measure of overall server efficiency, SWaP is calculated using the following
formula:
SWaP = Performance / (Space * Power Consumption) where,
• Performance is measured by industry-standard audited benchmarks such as those sponsored by the Systems
Performance Evaluation Corporation (SPEC)
• Space refers to the height of the server in rack units (RUs)
• Power is measured by watts used by the system, taken during actual benchmark runs or from vendor’s site
planning guides
8-Throughput Computing Realized
Sun Microsystems, Inc.
Table 3 provides an example comparison made using the SWaP metric, comparing the Sun Fire T2000 server to an
IBM eSeries Xseries x346 server running the SPECWeb2005 benchmark1. As measured by the SWaP metric, the Sun
Fire T2000 server exceeds this competitor by over four times.
Table 3. Comparison of a Sun Fire T2000 server with an IBM eServer Xseries x346 server using the SWaP metric
Feature
Sun Fire T2000 Server
IBM eServer x346
Sun Fire T2000 Server
Advantage
Space (RU)
2
2
None
Watts
330
438
25%
Performance (composite)
14,001
4,348
3.22 times
Performance/watt
42.427
9.927
4.27 times
SWaP
21.2
5.0
4.24 times
1.Disclosure: SPEC and SPECWeb are registered trademarks of the Standard Performance Evaluation Corporation. Sun Fire T2000
server (8 cores, 1 chip) 14001 SPECweb2005. IBM eServer Xseries x346 (2 cores, 2 chips) 4348 SPECWeb2005. Sun Fire T2000 server
results submitted to SPEC and other results from www.spec.org as of 12/06/05. IBM x346 specifications from product brochure, 09/
05/05: http://www-03.ibm.com/servers/eserver/xseries/x346.html. IBM x346 power rating estimated by calculating 70% of the
power supply data reported in the product brochure. Sun Fire T2000 server power consumption taken from measurements made
during the benchmark run.
9-The UltraSPARC T1 Processor with CoolThreads Technology
Sun Microsystems, Inc.
Chapter 2
The UltraSPARC T1 Processor with CoolThreads Technology
Dissatisfied with the performance characteristics of traditional single-threaded processors, Sun’s extensive inhouse design team—one of the largest microprocessor design engineering teams in the world—has taken a bold
new approach to processor design. Sun understands that the network-computing environments found in most
modern data centers are inherently multithreaded, where the execution speed of an individual thread is typically
less important than overall application throughput. For this reason, Sun is focusing on processors and
architectures that maximize throughput for commercial network-computing workloads. These efforts are resulting
in new chip multithreaded (CMT) processor technology that leverages the additional gains delivered by Moore’s
Law to provide thread-level parallelism rather than instruction-level parallelism.
Chip Multithreaded (CMT) Design
Recently, a number of processor vendors have started making multi-core processors available that place multiple
processors on a single die (so-called chip multi-processing). Unfortunately, many of these efforts too have resulted
in complex, hot, high-power processors that don’t solve the real issues driving processing and productivity in the
data center. The most significant potential benefits of multi-core technology derive from a fundamental
simplification and rethinking of how processors are designed and built.
Hardware Multithreading Processor Cores
Unlike traditional single-threaded processors and even most current multi-core processors, hardware
multithreaded processor cores allow rapid switching between active threads as other threads stall for memory.
Figure 4 illustrates a multithreaded processor core such as those found on the UltraSPARC T1 processor. The key to
this approach is that each core is designed to switch between up to four threads on each clock cycle. As a result,
the processor’s execution pipeline remains active doing real useful work, even as memory operations for stalled
threads continue in parallel.
C
Thread 4
CMT
Processor
Core
Thread 3
C
Thread 2
Thread 1
C
C
M
M
M
M
C
C
C
C
M
M
M
M
C
C
C
M
C
M
C
Compute time
M
Memory latency
M
M
Time
Figure 4. A hardware multithreaded processor core switches between a number of active threads, doing useful
work even while threads stall to perform memory-related operations
Hardware multithreading provides real value since it increases the ability of the execution pipeline to do actual
work. Utilization of the processor pipeline is greatly enhanced since a number of execution threads now share its
resources. The negative effects of memory latency are masked since the processor and memory subsystems remain
active in parallel to the processor execution pipeline.
10-The UltraSPARC T1 Processor with CoolThreads Technology
Sun Microsystems, Inc.
Chip Multithreading with CoolThreads Technology
Like single-threaded processors, multi-processing (multi-core) technology can also be used to scale and multiply
the benefits of hardware multithreading. Sun calls the result chip multithreading. Unlike complex single-threaded
processors, CMT processors utilize the available transistor budget to implement multiple hardware multithreaded
processor cores on a single silicon wafer or chip. Because these individual processor cores implement much simpler
pipelines (emphasizing TLP over ILP), they are also substantially cooler and require significantly less electrical
energy to operate. This innovative approach results in CoolThreads processor technology—multiple physical
instruction execution pipelines (one for each core), with several active thread contexts per pipeline or core.
Application throughput is greatly improved with CoolThreads processors, as is utilization of pipeline resources.
Thread-rich applications common in commercial workloads benefit greatly from this model, whether comprised of
larger multithreaded applications, or of large numbers of smaller single-threaded applications. The number of
simultaneous threads that can be accommodated is quite large, and a wide range of processor designs are
possible. Figure 5 illustrates a four-core UltraSPARC T1 processor in which four hardware multithreaded cores are
combined into a single CMT processor, supporting up to sixteen active execution threads.
Thread 16
C
Thread 15
C
Thread 14
Thread 13
C
C
M
Thread 10
C
C
M
Thread 7
C
Thread 6
Thread 5
C
C
M
M
C
C
C
Thread 3
C
Thread 2
C
C
M
M
M
C
C
M
M
M
M
C
C
C
C
M
M
M
M
Memory latency
M
C
C
C
M
M
C
C
C
Compute time
M
M
M
C
M
M
M
M
C
C
C
C
C
C
M
M
M
M
M
M
M
M
M
M
Thread 4
Thread 1
C
C
C
C
C
C
C
C
C
M
M
M
M
M
M
M
Thread 8
C
C
C
C
C
C
Thread 11
Thread 9
M
M
Thread 12
Four-core
CMT
Processor
M
M
M
M
M
Time
Figure 5. A four-core UltraSPARC T1 processor supporting four multithreaded cores and 16 active thread
contexts (an eight-core UltraSPARC T1 processor supports 32 active thread contexts)
11-The UltraSPARC T1 Processor with CoolThreads Technology
Sun Microsystems, Inc.
UltraSPARC T1 Processor Architecture
The UltraSPARC T1 multi-core, multithreaded processor is the first chip that fully implements Sun’s Throughput
Computing initiative. Each UltraSPARC T1 processor has either four, six, or eight cores, or individual execution
pipelines, all on the same chip—essentially an “SMP system on a chip”. Each core, in turn, supports up to four
hardware thread contexts, a set of registers that represent the thread's state. The processor is able to switch
threads on every clock cycle in a round robin ordered fashion, and skip threads that are stalled (e.g. those threads
waiting for a memory access). In spite of its innovative new technology, the UltraSPARC T1 processor is fully SPARC
v7, v8, and v9 compatible and binary compatible with earlier SPARC processors.
DDR-2 SDRAM
DDR-2 SDRAM
L2 cache
DDR-2 SDRAM
L2 cache
L2 cache
DDR-2 SDRAM
L2 cache
On-chip cross-bar interconnect
FPU
Core Core Core Core Core Core Core Core
0
1
2
3
4
5
6
7
System Interface
Buffer Switch Core
UltraSPARC T1 Processor
Bus
Figure 6. Block-level diagram of an eight-core UltraSPARC T1 processor
As shown in the Figure 6, the individual processor cores are connected by a high-speed, low-latency crossbar
interconnect implemented on the silicon itself. The UltraSPARC T1 processor includes very fast interconnects
between the processor, cores, memory, and system resources, including:
• A 134 GB/second crossbar switch that connects all cores
• A JBus interface with a 3.1 GB/second peak effective bandwidth
• Four DDR2 channels (25.6 GB/second total) for faster access to memory
12-The UltraSPARC T1 Processor with CoolThreads Technology
Sun Microsystems, Inc.
The memory subsystem of the UltraSPARC T1 processor is implemented as follows:
• Each core has an Instruction cache, a Data cache, an Instruction TLB, and a Data TLB, shared by the four thread
contexts. Each UltraSPARC T1 processor has a twelve-way associative unified Level 2 (L2) on-chip cache, and each
hardware thread context shares the entire L2 cache.
• This design results in unified memory latency from all cores (Unified Memory Access, UMA, not Non-Uniform
Memory Access, NUMA).
• Memory is located close to processor resources and four memory controllers provide very high bandwidth to
memory, with a theoretical maximum of 25GB per second.
• Extensive built-in RAS features include ECC protection of register files, Chipkill, memory sparing, soft error rates
and rate detection, and extensive parity/retry protection of caches.
Each core has a Modular Arithmetic Unit (MAU) that supports modular multiplication and exponentiation to help
accelerate Secure Sockets Layer (SSL) processing. There is a single Floating Point Unit (FPU) shared by all cores,
thus the UltraSPARC T1 processor is not an optimal choice for applications with floating point intensive
requirements.
Unsurpassed Reliability, Availability, and Serviceability
By its very nature, any sort of consolidation of business-critical applications or services must be coupled with
extremely high levels of reliability, availability, and serviceability (RAS). The innovative UltraSPARC T1 processor
together with the Sun Fire T1000 and T2000 servers and Solaris technology provide very high levels of reliability.
These capabilities stem not only from processor and system-level features, but from the very way that the
processor is designed.
UltraSPARC T1 Processor Technology
The UltraSPARC T1 processor itself provides a number of built-in features that contribute directly to greater
reliability and availability. These features include:
•
•
•
•
•
•
•
•
Extensive error correcting code (ECC) logic on the UltraSPARC T1 processor
Memory Chipkill support
Memory DRAM sparing/reconfiguration
Memory address parity protection
A memory hardware scrubber and an L2 cache scrubber
Soft error rates and soft error rate detection
DRAM channel deconfiguration
Redundancy (cache RAM row/column sparing, core sparing, and eFUSE)
The UltraSPARC T1 processor provides parity protection on its internal cache memories, including tag parity and
data parity on the D-cache and I-cache. The internal 3 MB L2 cache has parity protection on the tags and ECC
protection on the data. DIMMs employ ECC as well to help ensure high levels of data integrity. The system reports
and logs correctable ECC errors, and errors are corrected as soon as they are detected.
13-The UltraSPARC T1 Processor with CoolThreads Technology
Sun Microsystems, Inc.
Increased Reliability from CMT technology
Beyond individual features in the UltraSPARC T1 processor, CMT processor design itself can dramatically increase
reliability and availability by condensing many physical interfaces and components onto a single processor. Since
one UltraSPARC T1 processor essentially replaces many individual processors and interconnects in an equivalent
SMP system, the reliability effects can be profound. As shown in Figure 7, the associated consolidation of
individual processors and inter-chip communication is greatly simplified in a system based on the UltraSPARC T1
processor.
M M M M
M M M M
P P P P
Switch
P P P P
Switch
I/O
I/O
P P P P
Switch
Switch
M M M M
I/O
I/O
M M M M
P P P P
Switch
M M M M
M M M M
P P P P
Switch
P P P P
Switch
M M M M
M M M M
Memory
I/O
I/O
I/O
P P P P
Switch
Switch
I/O
P P P P
Switch
I/O
Traditional 32-thread SMP server: One cabinet,
eight processor/memory boards,
many switch ASICS and connections
P
P
P
P
P
P
P
P
Memory
Memory
Memory
Sun Fire Server with CoolThreads Technology:
One motherboard, no switch ASICS,
no board connections
Figure 7. By compressing 32 threads onto a single processor, the UltraSPARC T1 processor helps enable system
designs that are less susceptible to failure, with fewer parts and connections
The traditional SMP system with 32 single-threaded processors represented on the left of the figure is housed in a
refrigerator-sized cabinet and consists of eight processor/memory boards, each with four processors, memory, and
an I/O interface. Each board, in turn, has switch ASICS to connect the on-board components, and the cabinet has
another set of switch ASICS to connect one group of four boards to the other. In contrast, the Sun Fire server based
on the UltraSPARC T1 processor offers a much more integrated and tightly-coupled solution.
With the UltraSPARC T1 processor, a single chip now houses processors and interconnects, and without the need
for switch ASICS, the entire system now fits on a single motherboard and has many fewer parts and pins to fail.
Moreover, maintenance and servicing is reduced to a single board. In fact, Sun Fire T1000 and T2000 servers are
built using many fewer components than even competitive rack-mounted servers. Table 4 lists the component
counts of Sun Fire T1000 and T2000 servers as compared to several competitors. Smaller component counts
translate directly to larger numbers of years between service events.
14-The UltraSPARC T1 Processor with CoolThreads Technology
Sun Microsystems, Inc.
Table 4. Sun Fire T1000 and T2000 servers feature very low component counts when compared to competitors
System
Total number of parts
Sun Fire T1000 Server
2357
Sun Fire T2000 Server
3411
IBM p520
4064
HP DL585
5176
Dell 2850
5566
Dell 6850
7086
Protection of On-chip memories
As semiconductor technology continues to trend toward increasing chip densities, processors have inevitably
become more susceptible to soft error rates. These soft errors are typically caused by charged particles or radiation
and are transient and random in nature—usually doing no damage to the device. Many systems can tolerate some
level of soft errors. With these risks in mind, Sun systematically designed the UltraSPARC T1 processor with the
appropriate level of protection of its on-chip memories. In general, the UltraSPARC T1 processor protects memory
arrays with either single error correction/double error detection (SEC/DED) or parity protection. Redundant arrays
are protected with parity while non-redundant arrays are protected with ECC. Table 5 lists the UltraSPARC T1
processor’s on-chip memories and their corresponding protection mechanism.
Table 5. Protection mechanisms for UltraSPARC T1 on-chip memories
Memory Array
Protection
Integer Register File
ECC
Floating Point Register File
ECC
L1 Instruction Cache - Data
Parity/retry
L1 Instruction Cache - Tag
Parity/retry
Instruction TLB
Parity/retry
Data TLB
Parity/retry
L1 Data Cache - Data
Parity/retry
L1 Data Cache - Tag
Parity/retry
L2 Cache - Data
ECC
L2 Cache - Tag
ECC
A notable feature of this scheme is the ECC protection of the integer and floating point register files, an extensive
level of protection only matched by mainframe-class processors. While processor designs have mainly focused on
protecting the data path, caches, and main memories, the register file has largely been neglected. Because the
register file is accessed very frequently, the probability of errors is increased, and protecting the register file is
critically important. In addition, protecting the register file with ECC prevents errors from quickly spreading to
different parts of the system, and can help prevent application crashes or silent data corruption.
15-The UltraSPARC T1 Processor with CoolThreads Technology
Sun Microsystems, Inc.
Main memory reliability and availability
Standard ECC memory is a proven industry standard technology that has had a considerably positive impact on
server reliability. ECC memory is able to detect and correct single bit memory errors, comprising the vast majority
of memory errors. However, increased memory capacity, the density of memory on a single DIMM, and the higher
speed of memory subsystems have significantly increased the risks of multi-bit memory errors that cannot be
corrected by standard ECC memory and result in system hangs. To address this issue, designers implemented
Chipkill technology to correct multi-bit memory errors, increasing system availability considerably.
The UltraSPARC T1 processor protects main memory using several mechanisms. Chipkill technology is used to
withstand multi-bit memory errors within a DRAM device, including a failure that causes incorrect data on all data
bits of the device. The Chipkill mechanism in UltraSPARC T1 processors uses Galois Field instead of Hamming in its
Chipkill implementation. The Galois Field algorithm provides higher bandwidth than Hamming Chipkill (21.33 vs.
10.66 GB/second) and it can correct any error contained within a single memory nibble (4 bits), while detecting any
uncorrectable errors contained within any two nibbles. When writing data to the DIMM, data is written in the form
of a checksum appended to the data. If a single-nibble memory error occurs, then the data is immediately
recovered by recalculating the data from the checksum information. This procedure allows the system to correct
not only the single-bit errors that standard ECC memory can correct but also 2, 3, and 4-bit errors and even a whole
DRAM chip failure. Even assuming larger size memories that increase the likelihood of systems to crash due to a
memory error, Chipkill-equipped servers fail at a rate of about two orders of magnitude less than those with an
ECC-protected subsystem.
In conjunction with Chipkill, DRAM sparing is implemented in the UltraSPARC T1 processor to improve main
memory availability. Where Chipkill detects a failed DRAM chip, DRAM sparing reconfigures a DRAM channel to
map out the failed chip, effectively replacing it with a corrected DRAM chip. This technique restores the capability
of correcting any random single-nibble error and allows the system to run with minor impaired memory error
protection until the DIMM can be replaced.
System-level RAS features
In addition to RAS features in the UltraSPARC T1 processor, Sun Fire T1000 and T2000 servers implement a variety of
technologies to improve reliability.
• Hot-swappable and hot-pluggable components
Sun Fire T2000 server hardware is designed to support hot-swapping of both the fan units and power supplies. By
using the proper software commands, these components can be installed or removed while the system is
running. Hot-swap technology significantly increases the system’s serviceability and availability by providing the
ability to replace fan units and power supplies without service disruption.
The Sun Fire T2000 server features two hot-swappable power supplies so the system can continue to operate if
one of the power supplies were to fail or if one power source were to fail. The server also has a single hotswappable blower unit that works in conjunction with the power supply fans to provide cooling for the internal
disk drives. If the blower unit fails, the power supply unit fans provide sufficient cooling to keep the disk
subsystem running until the blower can be replaced. The Sun Fire T2000 server also features three hotswappable system fans. The fans enable the system to continue operating with adequate cooling in the event
that one of the fans fails. Four hot-pluggable disk drives are also supported in the Sun Fire T2000 server.
16-The UltraSPARC T1 Processor with CoolThreads Technology
Sun Microsystems, Inc.
• Environmental Monitoring
Sun Fire T1000 and T2000 servers feature an environmental monitoring subsystem designed to protect the server
and its component against extreme temperatures, lack of adequate airflow, power supply failures, and
hardware faults. Temperature sensors are located throughout the system to monitor the ambient temperature
of the system and internal components. The server software and hardware help ensure that the temperatures
within the enclosure do not exceed safe operational ranges.
All error and warning messages are sent to the system controller system console and are logged in the Advanced
Lights Out Monitor (ALOM) console log file. Required LEDs remain lit after an automatic system shutdown
occurs, to aid in problem diagnosis. The power subsystem is handled in a similar fashion by monitoring power
supplies and reporting any fault through the front and rear panel LEDs.
• Fault Management and Predictive Self-Healing
Sun Fire T1000 and T2000 servers feature the latest fault management technologies, including Fault
Management and Predictive Self-Healing introduced with the Solaris 10 OS. Self-healing technology enables Sun
systems to accurately predict component failures and to mitigate many serious problems before they actually
occur. This technology is incorporated into both the hardware and software of the Sun Fire T1000 and T2000
servers. More details on Solaris Fault Management and Predictive Self Healing are provided later in this
document.
• Automatic System Recovery
Sun Fire T1000 and T2000 servers provide for automatic system recovery (ASR) from failures in UltraSPARC T1
processor cores, memory modules, or PCI cards. The ASR functionality allows the system to resume operation
after experiencing certain nonfatal hardware faults or failures. When ASR is enabled, the system’s firmware
diagnostics automatically detect failed hardware components. An auto-configuring capability designed into the
system firmware enables the system to unconfigure failed components and to restore system operation. As long
as the system is capable of operating without the failed component, the ASR features allows the system to
reboot automatically, without operator intervention.
17-Sun Fire T1000 and T2000 Server Architecture
Sun Microsystems, Inc.
Chapter 3
Sun Fire T1000 and T2000 Server Architecture
Both the Sun Fire T1000 server and the Sun Fire 2000 server have been designed to provide breakthrough
performance while maximizing reliability and minimizing power consumption and complexity. This section details
the physical and architectural aspects of these systems.
Sun Fire T1000 Server Architecture Overview
The compact Sun Fire T1000 server is an ideal platform for the delivery of horizontally-scaled transaction and web
services. The server is designed to address the challenges of today's data center by delivering an excellent price-toperformance ratio with greatly reduced power consumption and a small physical footprint. Depending on the
model selected, the Sun Fire T1000 server features a single six- or eight-core UltraSPARC T1 processor.
Enclosure
The 1U Sun Fire T1000 server enclosure is designed for use in a standard 19-inch rack (Table 6). The chassis consists
of a base unit, a removable top cover, and a front bezel.
Table 6. Dimensions and weight of the Sun Fire T1000 server
Dimension
U.S.
International
Height
1.75 inches (1 RU)
4.3 centimeters
Width
16.8 inches
42.5 centimeters
Depth
19.0 inches
48.3 centimeters
Weight (with side rails)
24 pounds
10.9 kilograms
The Sun Fire T1000 server includes the following major components:
•
•
•
•
•
•
An UltraSPARC T1 processor with six or eight cores
Up to 16 GB of DDR2 SDRAM memory in 8 available memory slots
Four on-board Gigabit Ethernet ports
One PCI Express (PCI-E) slots
One SATA disk drive
Advance Lights out Management (ALOM) system controller
System Board
The Sun Fire T1000 server system board contains all of the logic components for the system (Figure 8). These
components include the UltraSPARC T1 processor and all DIMMS, the I/O bridge ASIC, the I/O subsystem, and the
service processor subsystem. The system board also contains the rear-panel I/O connectors for the host system and
the service processor as well as a PCI Express (PCI-E) expansion card connector. These components are described
below.
18-Sun Fire T1000 and T2000 Server Architecture
Sun Microsystems, Inc.
DDR-2 SDRAM
DDR-2 SDRAM
UltraSPARC T1 Processor
[ALOM/host
I2C devices]
JBus
SSI
Disk
OnChip
DRAM
TPM
ESP
FPGA
Service SSP Bus
Processor
x4 PCI-E
Host
OBP
Flash
XBus
FPGA
Flash
DUART
DVD
Enet
Mgmt.
Console
DVD
Serial
Mgmt.
Console
RJ-45
RJ-45
Serial
Port (DB9)
PCI-X
Dual
GBE
+Bridge
PCI-E Card Slot (x8)
ALOM
Flash
I2C
I/O
ASIC
EBus
Dual
GBE
x8
Stacked
RJ-45
Stacked
RJ-45
SAS/
SATA
SATA HDD
10/100/1000 Base-T
Figure 8. Sun Fire T1000 server block level diagram
• Memory Subsystem
Each UltraSPARC T1 processor features four UltraSPARC DDR2 memory controllers, each with a 16-byte wide (128
data bits plus 16 ECC check bits) data interface. Two basic memory configurations are supported, with either two
or four DIMMs per controller. All DIMMs must have identical capacity. The Sun Fire T1000 server uses only two of
the four UltraSPARC T1 DDR2 memory controllers, with a maximum memory configuration of eight DIMMs.
DIMM sizes of 512 MB, 1 GB, and 2 GB are supported for a maximum memory configuration of 16 GB. DIMMs
must be installed four at a time (with one pair on each controller), yielding a four-DIMM configuration and an
eight-DIMM configuration.
• I/O Subsystem
The UltraSPARC T1 processor contains a JBus controller and an I/O ASIC is employed in the Sun Fire T1000 server
to translate between JBus and an industry-standard I/O bus. The I/O ASIC that performs this function in the
Sun Fire T1000 server contains two PCI Express (PCI-E) root complexes, each providing one PCI-E link. Each PCI-E
link supports a maximum port width or eight lanes. The two PCI-E interfaces operate independently of each
other and each supports link widths of one, two, four, or eight lanes. The JBus is clocked at 200 MHz.
The first PCI-E interface from the I/O ASIC connects directly to the system’s PCI-E expansion slot. Full link speed
and link width (x8) are supported. The second PCI-E interface connects to an on-board combination of an PCI-E to
PCI-X bridge and Ethernet controller, acting as the root of all integrated I/O within the system. The bridge/
Ethernet device, and the PCI-E link operate with an x4 link width.
19-Sun Fire T1000 and T2000 Server Architecture
Sun Microsystems, Inc.
• PCI-Express slot
The Sun Fire T1000 server provides one PCI express (PCI-E) slot for low profile cards. This slot supports 1x, 4x, and
8x link width cards.
• Network interfaces
The Sun Fire T1000 server includes four 10/100/1000Base-T auto-negotiating Ethernet ports, implemented with
two dual-ported controllers. Each of the four Ethernet RJ-45 connectors includes two LEDs; a green link indicator,
and an amber activity indicator. A fifth 10/100Base-T port is available on the back panel (Figure 9) for a remote
management connection as part of the ALOM system controller.
Storage
Internal hard drive storage is supported by a serial attached SCSI (SAS)/serial ATA (SATA) controller. The combined
SATA/SAS controller allows the Sun Fire T1000 server to provide a low-cost SATA configuration. The Sun Fire T1000
server supports a single internal hard drive. A diskless configuration is also supported. An expansion card is
necessary for the provision of external storage connectivity on the Sun Fire T1000 server.
Power Supply
The server includes a single 300W AC power supply. To provide for the best (i.e. lowest) possible power rating and
to minimize excess heat generation, high efficiency is a key aspect of the power supply design. The power supply
contains its own cooling fan that also provides cooling for the system’s hard disk drive. The power supply operates
from a wide range of 100 to 240V AC input.
Front and Back Panels
Figure 9 illustrates the front and rear panels of the Sun Fire T1000 server.
Indicators
Power supply Indicators
AC In
Serial port (ttya)
Indicators
10/100/1000
Ethernet ports
Figure 9. Sun Fire T1000 server, front and rear panels
PCI-Express slot
Network management port
Serial management port
20-Sun Fire T1000 and T2000 Server Architecture
Sun Microsystems, Inc.
External features of the Sun Fire T1000 server include:
• Front and rear indicator lights provide locator (white), service required (amber), and activity status (green) for
the system.
• Rear power-supply indicator lights convey the status of the power supply
• A single AC plug is provided on the rear of the system
• Four 10/100/1000Base-T autosensing Ethernet ports are provided
• A DB-9 TTYA serial port is provided for serial devices (not connect to the ALOM system controller serial port)
• Two management ports are provided for use with the ALOM system controller. The RJ-45 serial management
port provides the default connection to the ALOM controller. The network management port supports an
optional RJ-45/10/100Base-T connection to the ALOM system controller. The serial management port is always
available. The network management port is configured through the serial management port.
Sun Fire T2000 Server Architecture
The expandable Sun Fire T2000 Server is optimized to deliver transaction and web services, including J2EE
application services, enterprise application services (ERP, CRM, and SCM) and distributed databases. The Sun Fire
T2000 server is also an ideal platform for consolidated tier-1 workloads
Enclosure
The Sun Fire T2000 server features a compact, yet expandable 2U rack-mountable chassis (Table 7). The Sun Fire
T2000 server gives customers the flexibility to scale their processing and I/O needs without wasting precious space.
Table 7. Dimensions and weight of the Sun Fire T2000 servers
Server/Dimension
U.S.
International
Height
3.5 inches (2 RU)
8.9 centimeters
Width
17.3 inches
44.0 centimeters
Depth
24.3 inches
61.7 centimeters
Weight (without PCI cards or rack mounts)
37 pounds
17 kilograms
The Sun Fire T2000 server includes the following major components:
•
•
•
•
•
•
An UltraSPARC T1 processor with four, six, or eight cores
Up to 32 GB of DDR2 SDRAM memory in 16 available memory slots
Four on-board Gigabit Ethernet ports
Three PCI Express (PCI-E) and two PCI-X slots (one occupied by a disk controller)
Up to four hot-pluggable SAS disk drives
Advance Lights out Management (ALOM) system controller
System Board
Like the Sun Fire T1000 server, the Sun Fire T2000 server system board contains all of the logic components for the
system, including the UltraSPARC T1 processor and all DIMMs, the I/O bridge ASIC, the I/O subsystem, and the
service processor subsystem. The system board also contains the rear-panel I/O connectors for the host system and
the service processor. A logical block-level diagram of the Sun Fire T2000 system board is shown in Figure 10.
21-Sun Fire T1000 and T2000 Server Architecture
DDR-2 SDRAM
DDR-2 SDRAM
UltraSPARC T1 Processor
SSI
JBus
PCI-E
BUS A
JBus to PCI-E
Bridge
x8
x4
PCI-E
Switch
x4
PCI-E
x8
BUS B
Bus B
133/100 MHz, 64-bit
10/100/1000 Base-T
Quad SAS/SATA
Controller
PCI-X Card Slot
Stacked
RJ-45
RJ-45
RJ-45
Bus A
PCI-X Card Slot
Stacked
RJ-45
Dual
GBE
PCI-E Card Slot (x8)
Dual
GBE
DVD
Serial
Mgmt.
Console
PCI-E to PCI-X
x8
PCI-E Card Slot (x8)
PCI-E Card Slot (x8)
x8
DVD
Enet
Mgmt.
Console
x4
PCI-E
Switch
x8
ALOM Module
DDR-2 SDRAM
Southbridge
33 MHz,
32-bit
Posix Serial
Port (DB9)
ATA-100
DDR-2 SDRAM
Sun Microsystems, Inc.
Dual jack
USB
Dual jack
(Rear)
DVD
2.5-inch HDD
2.5-inch HDD
2.5-inch HDD
2.5-inch HDD
Figure 10. Sun Fire T2000 server block level diagram
• Memory
Unlike the Sun Fire T1000 server, the Sun Fire T2000 server uses all four of the available UltraSPARC T1 DDR2
memory controllers for a maximum memory configuration of 16 DIMMs. All DIMMs must have identical
capacity. DIMMs must be installed eight at a time (with one pair on each controller) yielding an 8-DIMM
configuration or a 16-DIMM configuration. DIMM sizes of 512 MB, 1 GB, and 2 GB are supported, resulting in a
maximum capacity of 32 GB for the system.
• I/O Subsystem
The Sun Fire 2000 server utilizes an I/O bridge ASIC that translates between the JBus interface on the UltraSPARC
T1 processor and an industry-standard I/O bus. On the Sun Fire T2000 server, two PCI-E root complexes are
provided, with each link having a maximum port width of eight lanes (x8). The two PCI-E interfaces operate
independently, and each supports link widths of one, two, four, or eight lanes. The JBus runs at 200 MHz while
the PCI-E links run at 2.5 GHz.
The three external x8 PCI-E expansion slots are connected to port A or B of the I/O ASIC through a pair of PCI-E
switches. Full link speed and link width are supported in each slot. The two PCI-X slots are connected to port B of
the I/O ASIC through a PCI-E switch and a PCI-E to PCI-X bridge. Port A devices share bandwidth with on-board
SAS/SATA and dual Gigabit Ethernet devices. Port B slots share bandwidth with an on-board dual Gigabit
Ethernet device.
22-Sun Fire T1000 and T2000 Server Architecture
Sun Microsystems, Inc.
• Expansion slots
The Sun Fire T2000 server is well equipped with both legacy PCI-X slots, and the more current PCI-E high-speed
slots. These slots can accommodate low-profile PCI cards that adhere to the MD2 physical specifications. Three
x8 PCI-E slots are provided in the Sun Fire T2000 server, and these also support x1 and x4 cards. Most cards,
including InfiniBand network cards, currently do not require more than an x4 slot for full-bandwidth operation.
Two PCI-X slots are provided on the Sun Fire T1000 server to support older PCI or PCI-X I/O cards. The PCI-X slots
are both 133 MHz, 64 bits wide, and 3.3V, providing the highest speed available in PCI-X configurations.
• Network interfaces
Like the Sun Fire T1000 server, the Sun Fire T2000 server includes four 10/100/1000Base-T auto-negotiating
Ethernet ports, implemented with two dual-ported controllers. Each of the four Ethernet RJ45 connectors
includes two LEDs, a green link indicator, and a yellow activity indicator. A fifth 10/100Base-T port is available on
the back panel (Figure 11) for remote management connection as part of the Advance Lights Out Management
(ALOM) controller.
• USB Ports
The Sun Fire T2000 server provides four USB 1.1 ports. Two of the ports are located on the left of the front panel
of the server and are associated with one USB controller. The second two USB ports are located on the right side
of the rear of the panel and are associated with a second USB controller.
Storage
The Sun Fire T2000 server supports up to four hot-pluggable hard disk drives, controlled by an LSI SAS1064
controller, providing four ports of SAS connectivity to the disk drives at bandwidths of 3 Gb/second full duplex for
each disk. The disks provided with the Sun Fire T2000 server are 73 GB, 10000-rpm SAS disks, 2.5-inch small form
factor server grade, and are certified for 24x7 operation. Due to the small physical size of these drives, a the high
spindle speeds, access times are extremely good. A side benefit of using small disks is that they allow designers to
maximize the air intake area at the front of the server to improve airflow, further increasing environmental
margins and server reliability.
An optional slimline DVD-R/CD-RW is also provided.
Power supplies
The Sun Fire T2000 server is equipped with a dual redundant hot-swappable power-supply system. One power
supply is sufficient to run a full-populated server. However, for maximum protection against power supply failures,
Sun recommends that both power supplies be installed in the system at all times. Both power supplies at First
Customer Ship are rated at 550 watts each, these will be replaced by 450 watt power supplies during early 2006. In
normal operation, the power supplies share the power demands of the system equally between the pair. The
power supplies can also be plugged into different electrical sources if dictated by availability requirements.
23-Sun Fire T1000 and T2000 Server Architecture
Sun Microsystems, Inc.
Front and Back Panels
Figure 11 illustrates the front and back panels of the Sun Fire T2000 server.
Indicators and buttons
DVD-R/CD-RW
USB ports (2, 3)
PCI-Express slot (0)
Management ports
(serial and network)
Blower unit
Power supply 0
Disk drives
Power supply 1
Indicators
Serial port (ttya)
PCI-Express slot (1, 2)
GBE Ports
PCI-X slot (0, 1)
USB ports (0, 1)
Figure 11. Sun Fire T2000 server, front and rear panels
External features of the Sun Fire T2000 server include:
•
•
•
•
•
•
•
•
•
•
•
Four USB ports are provided, two on the front panel, and two on the rear.
One slimline DVD-R/CD-RW is accessed through the front panel.
Up to four hot-pluggable 73 GB SFF SAS disks are supported
Front and rear indicator lights provide locator (white), service required (amber), and activity status (green) for
the system.
Rear indicator lights on each power supply convey the status of the power supply.
A single AC plug is provided on each hot-swappable power supply
Three hot-swappable fans (not shown) can be accessed through a panel on the top of the enclosure. The hotswappable blower unit is accessed through the back panel, while the system is running.
Four 10/100/1000Base-T autosensing Ethernet ports are provided
A DB-9 TTYA serial port is provided for serial devices (not connected to the ALOM system controller serial port)
Two management ports are provided for use with the ALOM system controller. The RJ-45 serial management port
provides the default connection to the ALOM controller. The network management port supports an optional
RJ-45/10/100Base-T connection to the ALOM system controller. The serial management port is always available.
The network management port is configured through the serial management port.
Three PCI-E slots and two PCI-X slots are accessed through the rear panel.
24-Getting the Most from Throughput Computing: A Complete and Cohesive Software and Management Environ-
Sun Microsystems, Inc.
Chapter 4
Getting the Most from Throughput Computing:
A Complete and Cohesive Software and Management Environment
As the industry has witnessed many times, having a fast processor or even well-designed system is often not
sufficient for success. New technology often requires time for tools and applications to arrive, and delivering agile
and highly-available services that take advantage of available resources requires stable development tools,
operating systems, middleware and management software. Fortunately, in spite of the breakthrough
UltraSPARC T1 processor technology, Sun Fire T1000 and T2000 servers provide full binary compatibility with earlier
SPARC systems and are delivered ready to run with pre-loaded tools and the solid foundation of the Solaris OS.
The Solaris Operating System: Scalability and Support for CoolThreads Technology
The Solaris 10 Operating System is specifically designed to deliver the considerable resources of UltraSPARC T1
processor based systems such as the Sun Fire T1000 and T2000 servers. In fact, the Solaris 10 OS provides new
functionality for optimal utilization, relentless availability, unparalleled security, and extreme performance for
both vertically and horizontally scaled environments. The Solaris 10 OS runs on a broad range of SPARC and x86based systems and compatibility with existing applications is guaranteed.
The Solaris 10 OS provides specific features that facilitate Throughput Computing. One of the most attractive
features of systems based on the UltraSPARC T1 processor is that they appear as a familiar SMP system to the
Solaris OS and the applications it supports. The Solaris 10 OS has incorporated many features to improve
application performance on CMT architectures:
• CMT awareness
The Solaris 10 OS is aware of the UltraSPARC T1 processor hierarchy so that the scheduler can effectively balance
the load across all the available pipelines. Even though it exposes the UltraSPARC T1 processor as 32 logical
processors, the Solaris OS understands the correlation between cores and the threads they support.
• Fine-granularity manageability
The Solaris 10 OS has the ability to enable or disable individual processors. In the case of the UltraSPARC T1
processor, this ability extends to enabling or disabling individual cores and logical processors. In addition,
standard Solaris OS features such as processor sets provide the ability to define a group of logical processors and
schedule processes or threads on them.
• Binding interfaces
The Solaris OS allows considerable flexibility in that processes and individual threads can be bound to either a
processor or a processor set, if required or desired.
• Solaris Containers
Comprised of several key technologies, Solaris Containers provide fine-grained partitioning, virtualization, and
allocation of resources within a given Solaris instance. For example, the resources of a single UltraSPARC T1
processor in a Sun Fire T1000 or T2000 server can be easily partitioned into multiple containers, with each
securely supporting a separate web or application server.
25-Getting the Most from Throughput Computing: A Complete and Cohesive Software and Management Environ-
Sun Microsystems, Inc.
Solaris Containers Technology
Solaris Containers consist of a group of technologies that work together to efficiently manage system resources,
virtualize the environment, and provide a complete, isolated, and secure runtime environment for applications.
Solaris containers include important technologies that work together with the fair-share scheduler: Solaris Zones
partitioning technology and resource management tools. Solaris Zones enable an administrator to create separate
environments for applications on a single system, while the resource management framework allows for the
allocation, management, and accounting of system resources such as CPU and memory.
• Solaris Zones
New to the Solaris 10 Operating System is a unique partitioning technology called Solaris Zones that can be
used to create an isolated and secure environment for running applications. A zone is a virtualized operating
system environment created within a single instance of the Solaris Operating System. Zones can be used to
isolate applications and processes from the rest of the system. This isolation helps enhance security and
reliability since processes in one zone are prevented from interfering with processes running in another zone.
• Resource Management
Resource management tools provided with the Solaris Operating System help enable system resources such as
CPU resources to be dedicated to specific applications. CPUs in a multiprocessor system can be logically
partitioned into processor sets and bound to a resource pool, which in turn can be assigned to a Solaris zone.
Resource pools provide the capability to separate workloads so that consumption of CPU resources do not
overlap, and also provide a persistent configuration mechanism for processor sets and scheduling class
assignment. In addition, the dynamic features of resource pools enable administrators to adjust system
resources in response to changing workload demands.
Fault Management and Predictive Self Healing
With the Solaris 10 OS, Sun introduced a new architecture for building and deploying systems and services capable
of fault management and predictive self-healing. Predictive Self Healing is an innovative capability in the Solaris 10
OS that automatically diagnoses, isolates, and recovers from many hardware and application faults. As a result,
business-critical applications and essential system services can continue uninterrupted in the event of software
failures, major hardware component failures, and even software mis-configuration problems.
• Solaris Fault Manager
The Solaris Fault Manager facility collects data relating to hardware and software errors. It automatically and
silently detects and diagnoses the underlying problem, with an extensible set of agents that automatically
respond by taking the faulty component offline. Easy-to-understand diagnostic messages link to articles in Sun’s
knowledge base to clearly guide administrators through corrective tasks that require human intervention. The
open design of the Solaris Fault Manager facility also permits administrators and field personnel to observe the
activities of the diagnostic system. With Solaris Fault Manager, the overall time from a fault condition, to
automated diagnosis, to any necessary human intervention is greatly reduced, increasing application uptime.
26-Getting the Most from Throughput Computing: A Complete and Cohesive Software and Management Environ-
Sun Microsystems, Inc.
• Solaris Service Manager
The Solaris Service Manager facility creates a standardized control mechanism for application services by
turning them into first-class objects that administrators can observe and manage in a uniform way. These
services can then be automatically restarted if they are accidentally terminated by an administrator, if they are
aborted as the result of a software programming error, or if they are interrupted by an underlying hardware
problem. In addition, the Solaris Service Manager software reduces system boot time by as much as 75 percent
by starting services in parallel according to their dependencies. An “undo” feature helps safeguard against
human errors by permitting easy change rollback. The Solaris Service Manager is also simple to deploy;
developers can convert most existing applications to take full advantage of Solaris Service Manager features by
simply adding a simple XML file to each application.
Predictive self healing and fault management provide the following specific capabilities on Sun Fire T1000 and
T2000 servers:
• CPU Offlining takes a core offline that has been deemed faulty. Offlined CPUs are stored in the resource cache
and stay offline on reboot unless the processor has been replaced, in which case the CPU is cleared from the
resource cache (Note that CPU offlining on the Sun Fire T1000 and T2000 server will be delivered after their
initial release).
• Memory Page Retirement retires pages of memory that have been marked as faulty. Pages are stored in the
resource cache and stay retired on reboot unless the offending DIMM has been replaced, in which case affected
pages are cleared from the resource cache.
• I/O Retirement logs errors and faults.
• fmlog logs faults detected by the system.
The Java Enterprise System (JES)
The software industry has traditionally offered point products that solve specific parts of a problem, leaving it to
customers to integrate those products into a solution that can support their business applications. Organizations
don’t purchase their operating systems by assembling core components such as drivers, schedulers, command,
and administration utilities and it doesn’t make sense for them to assemble and integrate traditional middleware
this way either.
Suns Java Enterprise System (JES) provides a complete set of infrastructure software that is integrated to work as a
whole, and that offers shared components, common technologies, a consistent architecture and user experience.
Using world-class software, Sun redefines the software system from the operating system up through the J2EE
specification layer. Customers can write their business applications to Java software standards, leverage Java
Enterprise System network services, and Sun delivers the end-to-end solution to run them.
Bundled with the Sun Fire T1000 and T2000 servers, and now available at no cost, the Java Enterprise System
2005Q1 release includes the components listed in Table 8.
27-Getting the Most from Throughput Computing: A Complete and Cohesive Software and Management Environ-
Sun Microsystems, Inc.
Table 8. Java Enterprise System categories and components
Category
Components
Network Identity Services
Sun Java System Directory Proxy Server 5 2005Q1
Sun Java System Directory Server 5 2005Q1
Sun Java System Access Manager 6 2005Q1
Web and Application Services
Sun Java System Application Server Enterprise Edition 8.1 2005Q1
Sun Java System Message Queue 3 Enterprise Edition 2005Q1
Sun Java System Web Server 6.1 SP4 2005Q1
Portal Services
Sun Java System Portal Server 6 2005Q1
Sun Java System Portal Server Mobile Access 6 2005Q1
Sun Java System Portal Server Secure Remote Access 6 2005Q1
Communication and Collaboration
Services
Sun Java System Calendar Server 6 2005Q1
Sun Java System Instant Messaging 7 2005Q1
Sun Java System Messaging Server 6 2005Q1
Availability Servicesa
Sun Cluster 3.1 9/04
Sun Cluster Agents for JES Services (Web, application, directory, messaging, and
calendar servers and message queues)
a. As of this writing, Sun Cluster software is not supported on the Sun Fire T1000 and T2000 servers
Sun Studio 11 Compilers and Analysis Tools
Sun Studio 11 software is provided on each Sun Fire T1000 and T2000 server, offering the latest release of recordsetting optimizing compilers and tools for the C, C++, and FORTRAN developer. These compilers deliver the highest
optimizations and the best performance in the development of 32-bit and 64-bit applications on Sun’s newest
hardware platforms, including Sun Fire T1000 and T2000 servers as well as the latest multi-core x64 and x86
platforms. Sun Studio now removes the price barrier and is available at no cost.
Sun Studio software compilers allow developers to leverage the latest in parallel programming and maximize
throughput on multi-core and multi-threaded systems. In addition, even single-threaded applications gain as the
compilers can identify parallelization opportunities—generating back-end code that automatically parallelizes
execution without source-code changes. Integrated into a fully-featured IDE, Sun Studio 11 software includes an
advanced graphical debugger and a suite of performance analysis tools that simplify the development process for
even the most sophisticated programming needs.
• Acceleration for multi-core, multi-processor, and multithreaded technology
Sun Studio software can help achieve higher system throughput with multithreaded applications. Applications
can be built using the OpenMP v2.5 application interface (API) with C, C++, and FORTRAN code, along with an
improved debugger and performance analysis tools. The OpenMP implementation supports nested parallelism,
where nested regions can be executed by a team of two or more threads. Paired together with multi-core and
CMT optimizations, applications can maximize performance by taking advantage of the latest UltraSPARC T1,
x86, and x64 based systems.
• Highest performance on Sun platforms
Optimizing compilers produce record-setting runtime performance that consistently exceeds performance from
open-source alternatives and prior Sun Studio software releases.
28-Getting the Most from Throughput Computing: A Complete and Cohesive Software and Management Environ-
Sun Microsystems, Inc.
• Simple debugging
An intuitive graphical user interface makes it easy to access advanced debugging features. Breakpoints can be
set, variables examined, and the call stack navigated—all via the debugger’s convenient menus and buttons.
The unique Fix and Continue feature lets developers slash turnaround time for fixes and achieve greater
debugging productivity. Developers can even debug multi-threaded OpenMP code as well as seamlessly debug
code in mixed languages, including C, C++, FORTRAN, and Java.
• Sophisticated performance analysis tools
Developers can assess the performance of their applications, from algorithm changes to hardware system
counters, with one tool that correlates source code to actual machine execution. Also, new to this release,
dataspace profiling is available on UltraSPARC systems to provide unique views into the performance costs
associated with application memory references.
• Simple migration path
Source- and object-level compatibility with prior Sun Studio releases as well as GNU C/C++ compatibility features
simplify upgrade and adoption.
• Free license model
Sun Studio 11 introduces a new, no-charge licensing model with for-fee Sun Developer Support Plans. Standard
S4 support is available at launch with new Sun Developer Support Plans being introduced soon.
System Management
As the number of systems grow in any organization, the complexities of managing the infrastructure through its
lifecycle becomes increasingly difficult. Effective system management requires both integrated hardware that can
sense and modify the behavior of key system elements as well as advanced tools that can automate key
administrative tasks.
Advanced Lights-Out Management (ALOM) System Controller
The integral Advanced Lights Out Manager (ALOM) system controller allows Sun Fire T1000 and T2000 servers to be
remotely managed and administered. The ALOM software comes pre-installed on the Sun Fire T1000 server so
ALOM works as soon the administrator installs and applies power to the server. ALOM can then be customized to
work with a particular installation.
ALOM allows the administrator to monitor and control a server, either over a network or by using a dedicated serial
port for connection to a terminal or terminal server. ALOM provides a command-line interface that can be used to
remotely administer geographically-distributed or physically-inaccessible machines. In addition, ALOM allows
administrators to run diagnostics remotely (such as power-on self-test) that would otherwise require physical
proximity to the server serial port. ALOM can also be configured to send email alerts of hardware failures,
hardware warnings, and other events related to the server or to ALOM.
The ALOM circuitry runs independently of the server, using the server’s standby power. As a result, ALOM firmware
and software continue to function when the server operating system goes offline or when the server is powered off.
ALOM monitors disk drives, fans, CPUs, power supplies, system enclosure temperature, voltages, and the server
front panel, so that the administrator does not have to.
29-Getting the Most from Throughput Computing: A Complete and Cohesive Software and Management Environ-
Sun Microsystems, Inc.
ALOM specifically monitors the following Sun Fire T1000 and T2000 server components:
•
•
•
•
•
CPU temperature conditions
Enclosure thermal conditions
Fan speed and status
Power supply status
Voltage thresholds
Sun Management Center Software
Sun Management Center software is an element management system for monitoring and managing the Sun
environment. Sun Management Center software integrates with the leading enterprise management systems to
provide customers with a unified management infrastructure. The base package is free and provides hardware
monitoring. Advanced applications (add-ons) extend the monitoring capability of the base package. Sun
Management Center software provides:
•
•
•
•
•
•
•
Agents for managing Solaris OS (SPARC and x64/x86 platforms) and Linux operating systems
In-depth hardware and software diagnostics
Aggregate CPU utilization reporting
Event and alarm management for thousands of attributes
Corrective action automation through scripts triggered by alarm thresholds
Secure management controls for remote dynamic reconfiguration
The ability to customize modules with a powerful, easy-to-use GUI
Sun Management Center software version 3.6 supports the Sun Fire T1000 and T2000 servers.
Sun N1 System Manager
The Sun N1 System Manager is infrastructure lifecycle management software for deploying, monitoring,
patching, and managing large and small installations of Sun systems. Sun N1 System Manager takes a step-by-step
approach to unraveling the challenges of getting systems operational quickly:
• Discover
As systems are added to the management network, administrators can use Sun N1 System Manager to discover
bare metal systems based on a given subnet address or IP range.
• Group
Given the number of systems to manage and the constant re-purposing of systems, it is critical for IT
organizations to find ways to group resources together. Sun N1 System Manager enables users to logically group
systems together and perform actions across a group of systems as easily as performing actions on a single
system. Systems can be grouped by function (web servers versus grid computing), administrative responsibility,
or other categorization based on organizational needs.
• Provision
Sun N1 System Manager remotely installs operating systems (Solaris OS, RedHat, or SuSE Linux) onto selected
systems. Administrators can use this functionality to provision operating systems onto bare metal systems or
30-Getting the Most from Throughput Computing: A Complete and Cohesive Software and Management Environ-
Sun Microsystems, Inc.
reprovision existing systems. As the infrastructure life cycle continues, Sun N1 System Manager can update
firmware and provision software packages and patches to selected systems.
• Monitor
When systems are up and running, administrators can use Sun N1 System Manager to monitor system health,
helping to ensure that everything is running at the optimal levels. The software provides detailed hardware
monitoring for attributes such as fans, temperature, disk, and voltage usage, including bare metal systems. Sun
N1 System Manager also monitors OS attributes such as swap space, CPU, memory, and file systems.
Administrators can define specific threshold levels and set preferred notification methods, including e-mail,
pager, or Simple Network Management Protocol (SNMP) traps, for each monitored component as business
needs demand.
• Manage
Businesses require that infrastructure life cycle management extend beyond just deploying and monitoring
systems. Sun N1 System Manager includes Lights Out Management capabilities, such as powering systems on
and off, and remote serial console access to help IT organizations manage their IT infrastructure from remote
locations. Leveraging Sun N1 System Manager software’s Role-Based Access Control (RBAC) feature,
organizations can grant permissions to specific users to perform specific management tasks.
• Hybrid User Interface
Sun N1 System Manager offers users a hybrid user interface (UI), accessible from the Web, that integrates both
the GUI and CLI into one console. With this hybrid UI, operations performed in the GUI are simultaneously
reflected in the CLI, and vice versa.
The Sun N1 System Manager, version 1.2 provides support for the Sun Fire T1000 and T2000 servers.
31-Conclusion
Sun Microsystems, Inc.
Chapter 5
Conclusion
Delivering the benefits of Throughput Computing in the Participation Age will require a comprehensive approach
that includes innovative processors, system platforms, operating systems, along with application, middleware, and
management technology. With its strong technology positions and R&D investments in all of these areas, Sun is in
a unique position to deliver on this vision. Far from futuristic, Sun has effective solutions today that can help
organizations cope with the need for performance and capacity while effectively managing space, power and heat.
Powered by the breakthrough UltraSPARC T1 processor, Sun Fire servers with CoolThreads technology represent a
powerful new approach that delivers virtually unprecedented levels of performance while fundamentally changing
the equation on power and cooling. The result is data center infrastructure that can scale to meet new challenges,
even as it allows the organization to act responsibly toward the environment and the bottom line.
As the first systems to fully implement Sun's Throughput Computing vision, Sun Fire T1000 and T2000 servers
provide the resources for considerable web, application, and database consolidation while preserving customers
investments in SPARC/Solaris technology. With innovations such as Solaris Containers and Java technology,
customers can move to adopt this radical new technology without disrupting their ongoing operations. Innovative
Java Enterprise System middleware and Sun N1 system management software round out these offerings, providing
a complete and compelling solution.
sun.com
Copyright © 2005 Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, California 95054, U.S.A.
All rights reserved.
This product or document is protected by copyright and distributed under licenses restricting its use, copying, distribution, and decompilation. No part of this product or
document may be reproduced in any form by any means without prior written authorization of Sun and its licensors, if any. Third-party software, including font technology,
is copyrighted and licensed from Sun suppliers.
Sun, Sun Microsystems, the Sun logo, Sun Fire, CoolThreads, Solaris, Java, J2EE, and N1 are trademarks or registered trademarks of Sun Microsystems, Inc. in the U.S. and
other countries.
All SPARC trademarks are used under license and are trademarks or registered trademarks of SPARC International, Inc. in the U.S. and other countries. Products bearing
SPARC trademarks are based upon architecture developed by Sun Microsystems, Inc.
RESTRICTED RIGHTS: Use, duplication, or disclosure by the U.S. Government is subject to restrictions of FAR 52.227-14(g)(2)(6/87) and FAR 52.227-19(6/87), or DFAR 252.2277015(b)(6/95) and DFAR 227.7202-3(a). DOCUMENTATION IS PROVIDED AS IS AND ALL EXPRESS OR IMPLIED CONDITIONS, REPRESENTATIONS AND WARRANTIES, INCLUDING
ANY IMPLIED WARRANTY OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT, ARE DISCLAIMED, EXCEPT TO THE EXTENT THAT SUCH DISCLAIMERS HELD TO BE LEGALLY INVALID.
Sun Microsystems, Inc. 4150 Network Circle, Santa Clara, CA 95054 USA Phone 1-650-960-1300 or 1-800-555-9SUN Web sun.com
Download PDF
Similar pages