The Advantage of IBM Power Systems © 2009 IBM Corporation IBM Power Systems Power Systems continues a 7+ year run of growth Sun SPARC and HP/Itanium continue in decline UNIX Server Rolling Four Quarter Average Revenue Share Source: IDC Quarterly Server Tracker 2Q09 release, September 2009 2 © 2009 IBM Corporation IBM Power Systems What drives Power Systems growth? 1 2 3 3 Power Scale-up / out / within leadership Power Virtualization leadership SPARC, PA-RISC, Itanium and now x86 users are moving to Power © 2009 IBM Corporation IBM Power Systems 1 Power Scale-up, Scale-out and Scale-within leadership Over four decades of running the largest, most mission-critical applications Power Systems continue leadership in the primary requirements for large scale computing – Efficient Scalability – Performance – Reliability, Availability, and Serviceability – Manageability 8-64 core systems ARE Scale-up computing - where other x86 vendors are untested 4 © 2009 IBM Corporation IBM Power Systems Xeon 7500 will extend Nehalem architecture to four & eight socket systems “A MONSTER CHIP IS COMING. The next generation of MP processor is targeted for production later this year, and by all accounts it is going to be a monster. Nehalem-EX is part of the Nehalem family of processors, but compared to its siblings it has the highest cores/threads count, largest shared cache, highest CPU-to-CPU bandwidth, highest I/O bandwidth, highest memory capacity, highest memory bandwidth, greatest scalability, and highest level of Reliability/Availability/Serviceability. It’s expected to bring a gargantuan, unprecedented leap in capabilities and performance--the biggest leap in all of Xeon product history.” from a blog posted by Matt_K on Jun 8, 2009 5:45:18 PM available at www.intel.com Xeon 5500 vs Xeon 5400 per socket or per core Xeon 7500 vs Xeon 7400 per socket Database Transactions 2.5 2.5 Integer throughput 1.7 1.7 Floating point throughput 2.2 2.2 Memory 2.3 2.0 Memory Bandwidth 3.5 9.0 Comparison according to Intel Source: Intel Server Update, May 26, 2009 available at http://download.intel.com/pressroom/pdf/nehalem-ex.pdf 5 © 2009 IBM Corporation IBM Power Systems The 2007 570 is 28% faster than 2010 Xeon 5570 on TPC-C The Latest Power 570 5.0GHz system is even faster Virtualized Power performance beats Native Xeon For complete TPCC results, go to www.tpc.org tpmC 4.7GHz IBM Power 570 (8 chips, 16 cores, 32 threads) HP ProLiant DL370 G6 (2 chips, 8 cores, 16 threads) 6 Price / tpmC Data base Systems Availability 1,616,162 3.54 USD Enterprise 11/21/07 > 21 months ago 631,766 1.08 USD Standard 03/30/09 >4 months ago © 2009 IBM Corporation IBM Power Systems POWER6 beats Intel’s best on per core performance Database & Web application server licensing benefits from better per core performance 7 See slide “Substantiation for Power Systems Leadership Performance for detail © 2009 IBM Corporation IBM Power Systems Power Systems offer unmatched scalability >9.5 times the Xeon 5500 throughput for OLTP >4 times the Xeon 7500 throughput for integer 8 See slide “Substantiation for Power Systems Leadership Performance for detail © 2009 IBM Corporation IBM Power Systems It’s about the system – not just the chip Power Systems offers balanced systems design with the bandwidth to get the most performance and scalability from the processor 9 See slide “Substantiation for Power Systems Leadership Performance for detail © 2009 IBM Corporation IBM Power Systems Power Systems offer near-linear scalability Balanced systems design allows for linear performance as core-count and utilization increases 10 See slide “Substantiation for Power Systems Leadership Performance for detail © 2009 IBM Corporation IBM Power Systems ITIC Survey says Power Systems with AIX deliver 99.997% uptime - 54% of IT executives and managers say that they require 99.99% or better availability for their applications Power Systems with AIX delivers the best RAS of UNIX, Linux, Windows choices 1. Availability: The least amount of downtime 15 minutes a year 2.3 times better than the closest UNIX competitor more than 10X better than Windows 2. Reliability: The fewest unscheduled outages less than one outage per year 3. Serviceability: The fastest patch time 11 minutes to apply a patch Source: Network World, dated July 14, 2009, reports on the 2009 ITIC Global Server Hardware & Server OS Reliability Survey Results 11 © 2009 IBM Corporation IBM Power Systems So why would anyone buy an Itanium server? THE RECENT LAUNCH of Intel's Dunnington based six-core Xeon processors won't spell the end for the Itanium family. – Despite the significant performance gains of the new x86 based Xeon's, the company has confirmed that Itanium continues to be a viable choice for some customers primarily due to the 'Reliability – Availability – Serviceability' (RAS) features implemented in the VLIW based Itaniums. – Joachim Aertebjerg, Intel's Server Product Line director. quoted in “Dunnington won't sink Itanic says Intel”, By Ian Williams, Thursday, 18 September 2008, 09:11 Nehalem-EX will add new reliability, availability, and serviceability (RAS) features traditionally found in the company’s Intel® Itanium processor family, such as Machine Check Architecture (MCA) recovery. – Intel Previews Intel Xeon ‘Nehalem-EX’ Processor, May 26, 2009 Press Briefing 12 © 2009 IBM Corporation IBM Power Systems Power Systems RAS is designed for the toughest applications – from the same people who defined what “mainframe-class” means RAS Feature POWER6 SPARC Integrity Xeon Live Partition Mobility Yes No No Yes Live Application Mobility Yes No No No Partition Availability priority Yes No No No OS independent First Failure Data Capture Yes No No No Redundant System Interconnect No Yes No No Memory Keys Yes No No No Processor Instruction Retry Yes Yes No No Alternate Processor Recovery Yes No No No Dynamic Processor Deallocation Yes Yes Yes No Dynamic Processor Sparing Yes Yes2 Yes2 No Chipkill™ Yes Yes Yes Yes Redundant Memory Yes Yes Yes Yes Yes No No No Application/Partition RAS System RAS Processor RAS Memory RAS I/O RAS Extended Error Handling #1,2,3 - See “POWER6 RAS” in backup; See the following URLs for addition details:http://www-03.ibm.com/systems/migratetoibm/systems/power/availability.html http://www-03.ibm.com/systems/migratetoibm/systems/power/virtualization.html 13 © 2009 IBM Corporation IBM Power Systems Common management architecture and components to maximize resource utilization across the enterprise Integrated visibility, control & automation across heterogeneous business and technology assets Align IT operations with the business Govern and control the business Optimize the business Detailed platform & virtualization management of IBM systems Consolidated management across systems, networks and storage Integrated physical and virtual management across platforms Automated physical and virtual provisioning Platform Management Virtualization Management Service Management Integrated management to enable the delivery of critical business services 14 © 2009 IBM Corporation IBM Power Systems 1 Power Scale-up, Scale-out and Scale-within leadership Over four decades of running the largest, most mission-critical applications Power Systems continue leadership in the primary requirements for large scale computing – – – – Efficient Scalability Performance RAS Manageability * * * * >9.5 times the systems throughput of Xeon 5500 28% more OLTP throughput per core Best reliability, best availability, best serviceability The glue that lets you convert capability to service IBM Power Systems has proven experience in scale-up, scale-out and scale-within computing providing predictable, consistent performance. 15 © 2009 IBM Corporation IBM Power Systems 2 Power Virtualization leadership Celebrating 10 years of Power virtualization. Power Systems continue leadership in the primary requirements for consolidation – Choice of consolidating within an operating environment (OE) or consolidating multiple OEs – All applications run in a virtualized environment – Low overhead virtualization – Balanced scalability for a wide variety of applications – Manageability 1967 1973 1987 IBM announces physical partitioning IBM announces LPAR 1999 2004 2007 IBM announces LPAR on POWER™ IBM intro’s POWER Hypervisor™ IBM announces Live Partition Mobility IBM develops hypervisor for VM on mainframe x86 virtualization is developing and dependent on pieces from multiple vendors. 16 © 2009 IBM Corporation IBM Power Systems The challenge of scale out computing Typical serving running a single application is < 20% utilized Configuration planned for growth (20% unused?) What you pay for Configuration planned for peaks (50% unused?) System waits for I/O and memory access even when it is working (20% unused?) What you get Single workload on a single system – – Average Utilization: 20.7% Peak: 79% 80% of the hardware, software, maintenance, floor space, and energy that you pay for, is wasted 17 © 2009 IBM Corporation IBM Power Systems Reduce cost: Why is scalability important? The #1 reason IT managers deploy virtualization solutions is workload consolidation – Put simply, the more workloads that can be encapsulated within VMs and combined onto a single server, the higher the consolidation ratio and greater the cost reduction – The integrated combination of POWER architecture and PowerVM makes possible far higher consolidation ratios than the x86 architecture and VMware vSphere 18 © 2009 IBM Corporation IBM Power Systems IBM Confidential Reduce cost: PowerVM delivers superior scalability to maximize consolidation and drive down IT costs VMware ESX 3.5 VMware ESX 4.0 (in VMware Infrastructure 3) (in VMware vSphere 4) 4 8 64 64 GB 255 GB 4096 GB Virtual NICs per VM 4 10 256 CPUs per physical server 32 64 64 256 GB 1024 GB 4096 GB Scalability Factors Virtual CPUs per VM Memory per VM Memory per physical server PowerVM Source: http://www.vmware.com/files/pdf/key_features_vsphere.pdf 19 IBM Power Systems Reduce cost: Scalability of virtual CPUs VMware vSphere 4 – No more than 8 virtual CPUs can be assigned to a single VM (up from 4 in prior version) – The 8 virtual CPUs option is only available in the high-end Enterprise Plus edition – This constraint limits the type of high-end workloads that can be virtualized – Note: It does not matter if more than 8 CPU cores are available on the physical host (Example: a four-socket Nehalem EX x86 system will have 32 total cores, but a single VM cannot be configured to use all 32 of those cores) PowerVM – Can assign as many CPU cores as are available on the physical host (Example: a VM (LPAR) can use all 64 cores on a Power 595) – Each virtual CPU can run two threads, resulting in a maximum of 128 threads per VM – Result: A more effective solution for CPU-intensive workloads Source: http://www.vmware.com/files/pdf/key_features_vsphere.pdf 20 © 2009 IBM Corporation IBM Power Systems Improve service: PowerVM delivers superior flexibility to optimize IT resource utilization and boost responsiveness VMware ESX 3.5 VMware ESX 4.0 (in VMware Infrastructure 3) (in VMware vSphere 4) Dynamic virtual CPU changes in VM No Add (but not Remove) Yes Dynamic memory changes in VM No Add (but not Remove) Yes Dynamic I/O device changes in VM No No Yes Direct access to I/O devices from within VM No Some (with Nehalem) Yes Maximum simultaneous live migrations 4 4 8 Flexibility Factors PowerVM Source: http://www.vmware.com/files/pdf/key_features_vsphere.pdf 21 © 2009 IBM Corporation IBM Power Systems Every Power Systems benchmark published since July, 2004 has been run in a virtualized environment with the hypervisor active Over 70 leadership benchmarks published in last 5 years 22 © 2009 IBM Corporation IBM Power Systems Improve service: Consolidating diverse enterprise workloads VMware vSphere 4 – Only supports native x86-based workloads – mainly Windows and Linux/x86 – No plans to extend support to workloads created for other architectures – Management tool (vCenter) is limited to an x86-only subset of IT infrastructure – Perpetuates ‘silos of virtualization’ that require multiple management tools PowerVM – Supports all workloads built for AIX, IBM i and Linux (including Linux/x86) – IBM Systems Director can manage VMware, Xen, Hyper-V, KVM, PowerVM, and z/VM virtualized workloads with VMControl – Scales to support the most demanding mission-critical workloads Source: http://www.vmware.com/files/pdf/key_features_vsphere.pdf 23 © 2009 IBM Corporation IBM Power Systems Scale within simplifies the data center Server 0 Server 1 SCALE-OUT: Cables, routers, & switches are: additional points of failure difficult to keep track of expensive to maintain Server 2 Server 3 Server 4 SAN Switch 0 SAN Switch 1 Disk Farm 1 Disk Farm 2 Server 0 Part. 0 Part. 1 Part. 2 VIOS Partition 0 SCALE-WITHIN: Up to 90% reduction in cables, switch volume, adapters 24 Server 5 Server 6 Server 7 Server 1 Part. 3 Part. 4 Part. 5 Part. 6 Part. 7 VIOS Partition 1 SAN Switch 0 SAN Switch 1 Disk Farm 1 Disk Farm 2 © 2009 IBM Corporation IBM Power Systems 2 1. 2. 3. 4. Power Virtualization leadership Only PowerVM allows you to completely virtualize your datacenter Power Systems continue leadership in primary consolidation requirements: Support for “virtualize everything” including large production workloads Built in virtualization so you get the performance you expect Infrastructure designed for virtualization with superior bandwidth to support a wide variety of applications Multi-platform manageability to support Power, z, and x systems with a single management system A 40-year tradition continues 1967 1973 1987 IBM announces physical partitioning IBM announces LPAR 1999 2004 2007 IBM announces LPAR on POWER™ IBM intro’s POWER Hypervisor™ IBM announces Live Partition Mobility IBM develops hypervisor for VM on mainframe For over 10 years, IBM Power Systems has been fine-tuning highly integrated systems designed from the ground up for industrial strength virtualization. 25 © 2009 IBM Corporation IBM Power Systems 3 SPARC, PA-RISC, Itanium and now x86 users are moving to Power Clients are migrating to Power – Migration expertise – Dependable roadmap – Consolidation value As SPARC and Itanium decline, UNIX clients are likely to have two major choices - Power or x86. 26 © 2009 IBM Corporation IBM Power Systems Competitive migrations to IBM Power More than 1,750 IBM Migration Factory wins to date Competitive displacements 2009 momentum • Wins from Sun grew 111% QTQ • Wins from HP grew 44% QTQ 89% of migrations from Sun and HP (FY2006 through 1H2009) 27 © 2009 IBM Corporation IBM Power Systems Save up to 93% in annual energy costs! By consolidating 34 16-core Sun V890s into ONE rack of Power 570 systems --Reduce floor space required by 93% --Reduce processing cores by over 88% 34 Sun V890s (@ 20% utilization) 544 total cores @ 2.1 GHz Over 109 sq. ft. of floor space required up to 1,442 MWh annual energy One Rack of Power 570s (@ 60% utilization) 64 total cores @ 4.2 GHz Only 1 Rack – 7.6 sq. ft of floor space Save up to 1,344 MWh annually – up to $195k in energy savings per year! See “Power 570 power and efficiency claims” and “Power 570 consolidation claims*” charts in backup for full substantiation details. © 2009 IBM Corporation IBM Power Systems Consolidate up to 39 non-virtualized Sun SPARC Enterprise M5000 servers into one IBM Power 595 server 39 Sun SPARC Enterprise M5000 Servers* One IBM Power™ 595 server* • Reduce maximum energy use by up to 84% • Save up to 80% of the floor space 624 total cores @ 2.4 GHz Using an average of 20%maximum capacity 182,676 VA Maximum Power requirement** 64 total cores @ 5.0 GHz Using an average of 60% of maximum capacity 27,700 Maximum WATTS** See page “notes on 39 to 1” for detail © 2009 IBM Corporation IBM Power Systems Save up to $840,000 and 83% of the energy use! By consolidating on the Power 560 Express instead of the Sun M5000 -- Use 1/5 the rack space Coming From: 2 Racks: 13 V490 servers Maximum energy requirement of 22,750 WATTs List Price of $140,955 Maximum energy requirement of 2,400 WATTs 8U - One Power 560 Express server M5000 supports no more than 4 dynamic domains per system and would require Four M5000 servers to consolidate 13 V490s List price of $981,360 Maximum energy requirement of 14,952 WATTs 1 rack of 4 M5000 servers See Power 560 versus M5000 consolidation substantiation in backup for substantiation detail. If Solaris Containers could be used: Two M5000 servers required: List price of $490,680 and 7,296 Watts © 2009 IBM Corporation IBM Power Systems Energen Corporation reduces costs with migration from Sun to IBM Power Systems Client requirements Improve system performance and support for the company’s SAP ERP (enterprise resource planning) application by consolidating its sprawling 20-unit Sun server environment Reduce the total cost of ownership by cutting its licensing costs for the Oracle databases, which support the company’s SAP system Solution Migrated its SAP ERP system and Oracle databases onto two IBM Power servers [570s] Engaged IBM Business Partner Mainline Information Systems to demonstrate how leveraging virtualization technology could cut Oracle licensing costs Benefits Reduces Oracle licensing costs by 40 percent, contributing to US$500,000 in annual savings Provides a more efficient, available infrastructure that combines lower capital and operational costs with better performance and flexibility “We certainly made a saving on hardware costs, but the reduction in Oracle licensing costs was the main contributor to the total US$500,000 annual savings we achieved by migrating to IBM for our SAP software environment.” — Ron Payne, Director of Infrastructure Services Consumes significantly less floor space and power 31 © 2009 IBM Corporation IBM Power Systems SAP Application Servers are exploding as environments grow in complexity DEV Multiple Servers for each SAP System Landscape are required by SAP TEST QA SAP System 1 2001 1 landscape 59 batch jobs 400 users 32 2 landscapes upgrades/yr batch jobs users Multiple system landscapes per SAP functional solution Multiple operational stages per System 3 … 2003 2 2 124 850 PROD 2005 5 4 3 198 1400 landscapes parallel rollouts upgrades/yr batch jobs users 2006/7 9 6 8 310 2800 landscapes parallel rollouts upgrades/yr batch jobs users 2008+ 14 9 8 412 3100 landscapes parallel rollouts upgrades/yr batch jobs users © 2009 IBM Corporation IBM Power Systems Osram Sylvania consolidates SAP environment from Alpha & x86 to Power Business challenge: Replace 50 legacy HP Alpha and Windows servers running missioncritical SAP applications with a flexible, highly reliable system that would remain viable for more than five years and offer lower operational costs. Solution: Lowered operational costs and gained infrastructure flexibility when they teamed with IBM and SAP to migrate their SAP ERP applications to the IBM Power Systems platform •SAP ERP 6.0 •SAP NetWeaver® Business Information Warehouse 3.1 •IBM Power Systems models 570, 550 •IBM AIX® operating system •PowerHA for AIX •Oracle DB •IBM Global Business Services “We were convinced that IBM offered the best support for the transition, the best technology for operations, and the best strategy for long-term development.” Benefits: •Batch times reduced by a factor of five •User response times cut in half •Service to the business dramatically improved •Fewer servers means lower administration, maintenance, energy, cooling and license costs 33 Jeffrey Ruck Director of IT Infrastructure Services OSRAM SYLVANIA © 2009 IBM Corporation IBM Power Systems Internet Retail Innovation Supported by Managed Infrastructure Growth Business challenge: Move to a platform that would support Novell SUSE Linux applications and allow them to quickly scale up and stay one step ahead of the growing customer base Benefits: In same POWER™-based architecture footprint since 2005, scaling up to meet demand that has taken them to a projected US$1 billion in gross merchandise sales during 2008. Plans to use Live Partition Mobility on new POWER6 processor architecture for new application development and testing http://www.ibm.com/software/success/cssdb.nsf/CS/ARBN-7JZLCT?OpenDocument&Site=corp&cty=en_us 34 “Nothing performs like IBM Power as our database server. Best of all, our infrastructure remains simple – even as we add more processing capacity to meet growth.” Kris Ongbongan, Systems Manager, Zappos.com Power = Openness + Scalability © 2009 IBM Corporation IBM Power Systems 3 SPARC, PA-RISC, Itanium and x86 users are moving to Power Thousands are moving to Power 1. Migration expertise 2. Power Rewards to offset migration cost 3. Proven dependable roadmap 4. Proven utilization for maximum consolidation 5. The best choice for UNIX users Clients trust the migration experience of IBM and the proven capability of Power Systems to handle their toughest workloads. 35 © 2009 IBM Corporation IBM Power Systems IBM Power Systems are the right choice. 1 2 3 36 IBM is an expert in scale-up and scalewithin computing. IBM virtualization is years ahead of any x86 or UNIX alternative. UNIX clients trust IBM migration expertise and Power roadmaps. © 2009 IBM Corporation IBM Power Systems Substantiation for Power Systems Leadership Performance Sources for the biggest leap falls short – Comparison to Xeon 5500 and Xeon 7400 for integer throughput, floating point throughput and DB transactions based on SPECint_rate2006, SPECfp_rate2006, and TPC-C benchmark results shown on chart Power to Xeon substantiation. Comparison to Xeon 7500 based on Intel projections of Xeon 7500 to Xeon 7400. Note: This is not intended to be a projection of the benchmark results. Comparisons to Itanium based on best results of any HP Integrity system from “Compare UNIX Systems, Performance” at http://www03.ibm.com/systems/migratetoibm/systems/power/performance.html. All results are current as of August 3, 2009. Sources for unmatched scalability – Comparison to Xeon 5500 and Xeon 7400 for integer throughput, floating point throughput and DB transactions based on SPECint_rate2006, SPECfp_rate2006, and TPC-C benchmark results shown on chart Power to Xeon substantiation. Comparison to Xeon 7500 based on Intel projections of Xeon 7500 to Xeon 7400. Note: This is not intended to be a projection of the benchmark results. Comparisons to Itanium based on HP Integrity Superdome results from “Compare UNIX Systems, Performance” at http://www03.ibm.com/systems/migratetoibm/systems/power/performance.html. IBM results were for the IBM Power 595. All results are current as of August 3, 2009. Sources for Scalability is about systems – Source: HP QuickSpecs available at www.hp.com, Dell Datasheets and Dell™ PowerEdge™ Servers – PRESS KIT - Intel® Xeon® Processor 5500 Series available at www.intel.com and the POWER6 TechEx presentation. All data is current as of June 29, 2009. Sources for efficient linear scalability based on SAP SD 2-tier benchmark using SAP ERP release 6.0 (without the Unicode extensions). Benchmark detail and results are shown at “Compare UNIX Systems, Performance” at http://www03.ibm.com/systems/migratetoibm/systems/power/performance.html. All results are current as of August 3, 2009 37 © 2009 IBM Corporation IBM Power Systems Power to Xeon Substantiation Competitive benchmark results reflect results published as of May 26, 2009. The results are the best results for the systems compared. SPEC® and the benchmark names SPECrate®, SPECint®, and SPECjbb® are registered trademarks of the Standard Performance Evaluation Corporation. For the latest SPEC benchmark results, visit http://www.spec.org Competitive benchmark results reflect results published as of May 26, 2009. The results are the best results for the systems compared. For the latest TPC-C benchmark results, visit http://www.tpc.org 38 © 2009 IBM Corporation Power 570 power and efficiency claims IBM Power Systems Comparisons between the IBM Power 570, HP Integrity Superdome, HP Integrity rx8640, Sun SPARC Enterprise M8000, Sun Fire E6900 and Sun Fire V890. – All systems were compared based on maximum processor configurations unless recommended wattage was available for other configurations. Maximum configurations were used because that is the data point for which power requirements are defined. Other configurations of these systems could have different performance per watt metrics. – Performance/watt is calculated by dividing the performance metric by the recommended maximum power usage for site planning. Actual power used by the systems will be less than this value for all of the systems. – This information for the Power 570 is in "Model 9117-MMA server specifications" available at http://publib.boulder.ibm.com/infocenter/systems/scope/hw/topic/iphdx/sa76-0091.pdf. The power requirement for the Power 570 is 5600 watts. – The information for the HP Integrity Superdome is in “QuickSpecs HP Integrity Superdome Servers 16-processor, 32processor, and 64-processor Systems” available at http://h18000.www1.hp.com/products/quickspecs/Division/Division.html#11715. The power requirement for the 64 core Superdome is 12,196 watts. – The information for the rx8640 is in "QuickSpecs HP Integrity rx8640 Server" available at http://h18000.www1.hp.com/products/quickspecs/12471_div/12471_div.HTM. HP defines multiple maximum power ratings. This calculation uses the Marked Electrical for server which is consistent with the maximum selected for the other servers. The power requirement for the rx8640 is 5400 watts. – The information for the Sun SPARC Enterprise M8000 Server is in the "Sun SPARC Enterprise M8000 Server Site Planning Guide" available at http://docs.sun.com/source/819-4203-12/21ch3p.html. The power requirement for the M8000 is 10,500 watts. – The information for the Sun Fire E6900 Server is in the “Sun Fire E6900/E4900 Systems Site Planning Guide” available at http://docs.sun.com/source/817-4117-14/environment.html. The power requirement for the E6900 is 9,410 watts. – The information for the Sun Fire V890 Server is in the “Sun Fire™ V490/V890 Servers with UltraSPARC IV+ 2100MHz CPU/Memory Modules Supplement ” available at http://dlc.sun.com/pdf//820-0714-10/820-0714-10.pdf. The power requirement for the V890 is 4,843 watts. © 2009 IBM Corporation IBM Power Systems Power 570 consolidation claims * The virtualized system count and energy savings were derived from several factors: A performance factor of 5.67X was applied to the virtualization scenario based on SPEC® results source: www.spec.org . Power 570 (32-core, 16 chips, 2 cores per chip, 4.2 GHz) SPECjbb2005 1,390,087 bops, 86,880 bops/JVM as of 10/7/2008; Sun Fire V890 (16-core, 8 chips, 2 cores per chip) 2.1 GHz, SPECjbb2005 244,846 bops, 30,606 bops/JVM as of 9/25/2008. A virtualization factor of 3X was applied to the virtualization scenario using utilizations derived from studies conducted by Alinean available at http://www-935.ibm.com/services/us/cio/optimize/opt_wp_ibm_systemp.pdf. A factor of 2X was used to represent the ability to install two 32-core Power 570 systems in a single rack. Power consumption figures of 5600 W for the IBM Power 570 and 4843 W for the Sun Fire V890 were based on the maximum rates published by IBM and Sun Microsystems, respectively. Air conditioning power requirement estimated at 50% of system power requirement. Energy cost of $.0971 per kWh is based on 2008 YTD US Average Retail price to commercial customers per US DOE at http://www.eia.doe.gov/cneaf/electricity/epm/table5_6_b.html as of 9/25/2008. The reduction in floor space, power, cooling and software costs depends on the specific customer, environment, application requirements, and the consolidation potential. Actual numbers of virtualized systems supported will depend on workload levels for each replaced system. ** The virtualized system count and energy savings were derived from several factors: A performance factor of 1.7 was applied to the virtualization scenario based on SPEC® results source: www.spec.org . Power 570 (32-core, 16 chips, 2 cores per chip, 4.2 GHz) SPECjbb2005 1,390,087 bops, 86,880 bops/JVM as of 10/7/2008; Sun SPARC Enterprise M8000 (64-core, 16 chips, 4 cores per chip) 2.52 GHz, SPECjbb2005 817,158 bops, 51,072 bops/JVM as of 9/25/2008. A virtualization factor of 3X was applied to the virtualization scenario using utilizations derived from studies conducted by Alinean available at http://www-935.ibm.com/services/us/cio/optimize/opt_wp_ibm_systemp.pdf. A factor of 2X was used to represent the ability to install two 32-core Power 570 systems in a single rack. Power consumption figures of 5600 W for the IBM Power 570 and 10,500 W for the Sun M8000 were based on the maximum rates published by IBM and Sun Microsystems, respectively. Air conditioning power requirement estimated at 50% of system power requirement. Energy cost of $.0971 per kWh is based on 2008 YTD US Average Retail price to commercial customers per US DOE at http://www.eia.doe.gov/cneaf/electricity/epm/table5_6_b.html as of 9/25/2008. The reduction in floor space, power, cooling and software costs depends on the specific customer, environment, application requirements, and the consolidation potential. Actual numbers of virtualized systems supported will depend on workload levels for each replaced system. © 2009 IBM Corporation IBM Power Systems notes on 39 for 1 * The number of Sun SPARC Enterprise M5000 servers that a single IBM Power™ 595 server could replace was calculated based on SPECint_rate2006 results. The peak result for the M5000 is for a 2.4GHz system with 16 processors (chips) and 2 cores per chip. It has a result of 158. The M5000 result can be found at www.spec.org. It is current as of March 25, 2008. The IBM Power 595 server result is for a 5.0GHz system with 32 processor (chips) and 2 cores per chip. That result was submitted on April 8, 2008. It will also be posted on www.spec.org. It has a peak result of 2,080 users. Estimating cumulative capacity as the number of servers times the throughput result of a single server, the cumulative capacity of the 13 Sun SPARC Enterprise M5000 servers is 13 times 158 users or 2,054. The capacity of the single Power 595 server is greater than the cumulative capacity of the 13 M5000 servers. A virtualization factor of 3X was applied to the virtualization scenario using utilizations derived from studies conducted by Alinean available at http://www-935.ibm.com/services/us/cio/optimize/opt_wp_ibm_systemp.pdf. That is; the utilization rate for the non-virtualized capacity of the M5000 server is estimated to be 20% and the utilization rate for the virtualized capacity of the Power 595 is estimated to be 60%. The used M5000 capacity is therefore estimated as 39*158 * 20% = 1,232.4. The Power 595 server used capacity is estimated as 2,080* 60% =1,248. Therefore the capacity of the Power 595 server at 60% is > than the cumulative capacity of the 39 M5000 servers at 20% utilization . SPEC® and the benchmark names SPECrate®, SPECint®, and SPECjbb® are registered trademarks of the Standard Performance Evaluation Corporation. For the latest SPEC benchmark results, visit http://www.spec.org ** Sun SPARC Enterprise M5000 server Maximum AC power consumption of 4,684 VA was sourced from Sun SPARC Enterprise M4000/M5000 Servers Site Planning Guide available at http://docs.sun.com/source/819-2205-10/Chap2_environ.html as of March 25, 2008. The IBM Power 595 server maximum power requirement is 27,700 VA. The savings from using the Power 595 were calculated by multiplying the M5000 maximum by 39 for a total of 182,676VA. The Power 595 server maximum requirement of 27,700 VA is 15.16% of the 182,676. *** The Sun SPARC Enterprise M5000 is a rack system. The calculation of floor space here was based on using .1 of Sun Rack 1000-42 for each M5000. The dimensions of the Sun Rack 1000-42 are 23.5” wide x 39.4” deep. They were sourced for the Sun Rack 1000-42 Tech Specs available at http://www.sun.com/servers/rack/1000-42/specs.xml#anchor1 as of March 25, 2008. The IBM Power 595 is 30.5” wide x 58.5” deep for a system with up to 3 I/O drawers. The savings from using the Power 595 were calculated by multiplying the M5000 floor space by 39 for a total of 62.69 square feet. The © 2009 IBM Corporation square footage for the Power 595 is 12.39 square feet which is 19.76% of 62.69. IBM Power Systems Power 560 Performance and Efficiency Substantiation Substantiation: Notes: 1. Competitive benchmark results reflect results published as of September 12, 2008. The SPECint_rate2006 results can be found at www.spec.org. The Power 560 Express final publication will be Submitted on October, 7 2008. All systems were compared based on maximum processor configuration because that is the data point for which power requirements are defined. Other configurations of these systems could have better performance per WATT metrics. 2. SPEC® and the benchmark names SPECrate®, SPECint®, and SPECjbb® are registered trademarks of the Standard Performance Evaluation Corporation. Competitive benchmark results stated above reflect results published on www.spec.org as of October 6, 2008. The comparison presented above is based on the best performing 8-chip servers currently shipping by IBM, Sun, and HP respectively. For the latest SPEC benchmark results, visit http://www.spec.org. 3. SPECint_rate2006 Peak/core results are: IBM Power 560 Express with 8 chips and 16 cores and two threads per core with a projected result of 363. Sun Microsystems Sun SPARC Enterprise M5000 with 8 chips, 32 cores and 2 threads per core with a result of 264. HP Integrity rx7640 with 8 chips and 16 cores and 2 threads per core with a result of 201 4. Performance per Watt is calculated by dividing the performance by the maximum system power. 5. Space for the Power 560 is 8 rack units. The Sun SPARC Enterprise M5000 is 10 rack units. This information for the Power 560 is in "Model 8234-EMA server specifications" available at http://www-01.ibm.com/common/ssi/index.wss - search for Power 560. The information for the Sun SPARC Enterprise M5000 Server is in the "Sun SPARC Enterprise M5000 Servers Site Planning Guide" available at http://docs.sun.com/app/docs/coll/m5000-hw. HP integrity rx7640 is 10 rack units and specifications are available at http://h18000.www1.hp.com/products/quickspecs/12470_div/12470_div.PDF 6, Performance per watt is calculated by dividing the performance in the table above by the recommended maximum power for site planning. Actual power used by the systems will be •Source: http://www.spec.org/ •Power 560 results be submitted on Octoberfor 7,the 2008 less Express than thisPOWER6 value for all of thewill systems. This information Power 560 is in "Model 8234-EMA server specifications" available at http://www01.ibm.com/common/ssi/index.wss - search for Power 560. The power for the 560 is 2,400 WATTs. The information for the Sun SPARC Enterprise M5000 Server is in the "Sun SPARC Enterprise M5000 Servers Site Planning Guide" available at http://docs.sun.com/app/docs/coll/m5000-hw. The power requirement for the M5000 is 3,738 WATTS. HP © 2009 IBM Corporation integrity rx7640 is 2128 watts and specifications are available at http://h18000.www1.hp.com/products/quickspecs/12470_div/12470_div.PDF 42 PRELIMINARY Power 560 BENCHMARK RESULTS IBM Power Systems Power 560 Consolidation Substantiation Competitive benchmark results reflect results published as of October 6, 2008. The SPECint_rate2006 results can be found at www.spec.org. The Power 560 Express final publication will be Submitted on October, 7 2008. All systems were compared based on maximum processor configuration because that is the data point for which power requirements are defined. Other configurations of these systems could have better performance per WATT metrics. SPEC® and the benchmark names SPECrate®, SPECint®, and SPECjbb® are registered trademarks of the Standard Performance Evaluation Corporation. Competitive benchmark results stated above reflect results published on www.spec.org as of October 7, 2008. The comparison presented above is based on a consolidation of a legacy 8-core Sun UltraSPARC IV servers into a 16 core IBM Power 560. For the latest SPEC benchmark results, visit http://www.spec.org. SPECintRate_2006 Peak/core results are: POWER6: IBM Power 560 Express with 8 chips, and 16 cores @ 3.6 GHz and 2 threads per core with a projected result of 363. SPARC: Sun V490 with 4 chips, 8 cores @ 2.1GHz and 1 thread per core with a result of 78. *The virtualized system count and energy savings were derived from several factors: A performance factor of 4.6X was applied to the virtualization scenario based on SPECint_rate2006. Power 560 (16-core, 8 chips, 2 cores per chip, 3.6 GHz) 363, submitted on 10/07/2008; Sun Fire V490 (8-core, 4 chips, 2 cores per chip) 2.1 GHz, SPECint_rate2006 of 78. The performance factor is simply the SPECint_rate2006 result of the Power 560 Express divided by the result of the competitive Sun V490 server. A virtualization factor of 3X was applied to the virtualization scenario using utilization assumptions derived from an Alinean white paper on server consolidation. The tool assumes 19% utilization of existing servers and 60% utilization of new servers. Source - www.ibm.com/services/us/cio/optimize/opt_wp_ibm_systemp.pdf. Calculation Summary: the 560 to the Sun V490 performance ratio is 4.6 Multiply by 3 for the virtualization factor. Hence, 4.6 * 3 = 13.9 servers rounded to 13 V490 server can be consolidated into 1 560 server. The Sun V490 is 5U in height and 8 can fit into a 42U rack. The 560 is 8U in height and 5 560 systems can fit in a 42U rack. One 560 system is 16 cores per system. A Sun V490 has 8 cores per system. 13 systems multiplied by 8 cores is 104 cores. 92% more cores. Power consumption figures of 2400W for the IBM Power 560 and 1750W for the Sun Fire V490 were based on the maximum rates published by IBM and Sun Microsystems, respectively. This information for the 560 is in "Model 8234-EMA server specifications" available at http://www-01.ibm.com/common/ssi/index.wss - search for Power 560. Sun Fire V490 Maximum AC power consumption of 1750 WATTs was sourced from Sun Fire™ V490/V890 Servers with UltraSPARC IV+, 2100MHz CPU/Memory Modules Supplement available at http://dlc.sun.com/pdf/820-0714-10/820-0714-10.pdf as of September, 2008. 43 © 2009 IBM Corporation PRELIMINARY p550 BENCHMARK RESULTS IBM Power Systems Power 560 versus M5000 Consolidation Substantiation Power 560 Express Pricing: $140,955 Power 560 Express Server, Includes 16 Core 3.6 GHz POWER6 Processors 64GB System Memory, 4 x 146 GB SAS Disk Drives, 1 DVD-ROM, 2 Gb Ethernet Ports, and 4 Power Supplies (220 V with N+N Redundancy) Sun SPARC Enterprise M5000 pricing: $181,340 + $64,000 (64GB of memory) = $245,340 times 4 servers = $981,360 Sun Pricing: http://shop.sun.com/is-bin/INTERSHOP.enfinity/WFS/Sun_NorthAmerica-Sun_Store_US-Site/en_US/-/USD/ViewConfigurationsList;pgid=tyL4UHemJpNSR08nlpFb_str0000crh3TBti;sid=anhg_kXDZHdg_Q0QzxYo-6pe3_pCFlSyC9jXC_XKwbj_gYJOHk=?ProxyProductRefID=DUMMY3--HID-240460404@Sun_NorthAmericaSun_Store_US&CatalogCategoryID=hudIBe.dZb4AAAEUWEg5G_c2&ShowAllProducts=false Sun SPARC Enterprise M5000 Server, Includes 8 Quad-Core 2.4 GHz SPARC64 VII Processors, 4 CPU Board with 2 CPUs each 5 MB On Chip L2 Cache, and 64 GB System Memory (4 Memory Modules with 8 x 2 GB DDR2 DIMMs), 4 x 146 GB SAS Disk Drives, 1 DVD-ROM, 4 Gb Ethernet Ports, 2 I/O Trays with 4 PCI-e and 1 PCI-X Slots, 4 Power Supplies (110 V or 220 V with N+N Redundancy), RoHS-5 Compliant Quantity 4 SELX2B1Z $ 16,000.00 Sun SPARC Enterprise Server Memory Module, 8 x 2 GB DIMMs, 16 GB total memory, for SPARC Enterprise M4000 and M5000 servers, RoHS-5 Compliant Power Consumption: This information for the Power 560 is in "Model 8234-EMA server specifications" available at http://www01.ibm.com/common/ssi/index.wss - search for Power 560. The power for the 560 is 2,400 WATTs. The information for the Sun SPARC Enterprise M5000 Server is in the "Sun SPARC Enterprise M5000 Servers Site Planning Guide" available at http://docs.sun.com/app/docs/coll/m5000-hw. The power requirement for the M5000 is 3,738 WATTS. Actual power used by the systems will be less than this value for all of the systems. Four M5000 servers times 3,738 watts equals 14,952. 83% more power than one Power 560 at 2,400 Watts. 20% of Sun V490 SPECint_rate2006 of 78 is 15.6. 60% utilization of the SPARC enterprise M5000 using SPECintrate_2006 is 158.4. Hence, the M5000 using Solaris containers can support 10 Sun Fire V490 servers. It would require two M5000 servers to consolidate 13 V490s using Solaris containers. 44 © 2009 IBM Corporation IBM Power Systems # Cores GHz IBM System POWER Result Second Place Result POWER Faster By TPC-C 64-core 64 5 595 6,085,166 2,382,032 155% Fujitsu Primequest TPC-C 32-core 32 1.90 p5-595 1,601,784 1,354,086 18.2% Fujitsu P’Quest TPC-C 16-core 16 4.7 570 1,616,162 579,814 178.7% HP DL585 Benchmarks Second Place System (non-IBM) TPC-C 4-core 4 4.7 570 404,462 230,569 75.4% HP rx6600 SAP SD 3-tier Overall 32 1.90 p5-595 168,300 100,000 68.3% HP Superdome 64-core SAP SD 2-tier 16-core 16 4.7 570 8,000 4170 91.8% Sun T5240 SAP SD 2-tier 4-core 4 4.7 570 2,035 1,218 67.1% HP BL480c SAP SD 2-tier 2-core 2 2.10 p5-505 680 597 13.9% HP ProLiant ML370 3.6 GHz Oracle Apps Online 11.5.9 8 1.90 p5-570 15,004 DNP Oracle Apps. Std. Batch 11.5.9 8 1.90 p5-570 2,744,000 2,664,000 3.0% Fujitsu PrimePower 850 (16-core) SPECint_rate2000 4-core 4 2.10 p5-550 90.0 123 -26.8% Dell PowerEdge SPECfp_rate2000 4-core 4 2.10 p5-550 149 121 23.1% Sun Ultra 40 SPECint_rate2000 8-core 8 2.20 p5-575 200 200 0% SPECfp_rate2000 8-core 8 2.20 p5-575 382 214 78.5% SPECint_rate2000 16-core 16 1.90 p5-575 314 283 11% SPECfp_rate2000 16-core 16 1.90 p5-575 571 373 53.1% Bull NovaScale SPECint_rate2000 32-core 32 1.65 p5-590 529 537 -1.5% Fujitsu PrimePower 1500 SPECfp_rate2000 32-core 32 1.65 p5-590 870 766 13.6% Fujitsu Primequest 480 SPECint_rate2000 64-core 64 2.30 p5-595 1,513 1108 36.6% HP Superdome (1.6 GHz) SPECfp_rate2000 64-core 64 1.90 p5-595 2,406 1,257 91.4% SGI Altix 3000 SPECfp2006 1 5 595 24.9 16.9 47.3% HP rx6600 SPECint_rate2006 8-core 8 5 550 263 260 1.1% Sun X2270 SPECfp_rate2006 8-core 8 5 550 222 200 11% Fujitsu RX300 SPECsfs_R1.v3 SMP 8 2.20 p5-570 169,786 66,235 156.3% HP AlphaServer GS1280 SPECjbb2005 16-core 16 5 570 867,989 758,325 14.4% Tyan TX46 45 Dell PowerEdge/Fujitsu Primergy Sun X4600 Fujitsu PrimePower POWER vs. Best Competitive Result Comparing the best available results vs. POWER 64-core (32/64/128) IBM Power 595 TPC-C result of 6,085,166 tpmC, $2.81/tpmC, avail. 12/10/08 64-core (32/64/128) Fujitsu Primequest TPC-C result of 2,382,032 tpmC, $3.76/tpmC, avail. 12/04/08 32-core IBM p5-595 TPC-C result of 1,601,784 tpmC, $5.05/tpmC, avail. 04/20/05 32-core (16/32/64) Fujitsu P’Quest TPC-C result of 1,354,086 tpmC, $3.25/tpmC, avail. 11/22/08 16-core (8/16/32) IBM Power 570 TPC-C result of 1,616,162 tpmC, $3.54/tpmC, avail. 11/21/07 16-core (4/16/16) HP DL585 TPC-C result of 579,814 tpmC, $.96/tpmC, avail. 11/17/08 4-core (2/4/8) IBM Power 570 TPC-C result of 404,462 tpmC, $3.50/tpmC, avail. 11/26/07 4-core (2/4/8) HP rx6600 TPC-C result of 230,569 tpmC, $2.63/tpmC, avail. 12/01/06 Sources: http://www.spec.org http://www.tpc.org http://www.sap.com/benchmark/ http://performance.netlib.org/performance/html/PDSreports.html All results are as of 05/01/09 TPC-C results with processor chip/core/thread. SPEComp results: IBM cores = 2x chip, threads = 4x chip. SAP certification numbers can be found in SAP section of charts. Linpack results are SMP only. © 2009 IBM Corporation IBM Power Systems # Cores GHz IBM System POWER Result Second Place Result POWER Faster By Lotus NotesBench R6Mail 16 1.65 i5-595 175,000 120,000 45.8% 8 2-way HP ProLiant BL20p Lotus NotesBench D7 R6iNotes 16 1.8 p5-560Q 55,000 43,000 27.9% Sun T5120 SPEC OMPM2001 (peak) 2-core 2 3.8 JS12 12,885 7,612 69.2% Sun Fire X4200 SPEC OMPM2001 (peak) 4-core 4 4.2 520 20,443 13,817 47.9% Sun V40z SPEC OMPM2001 (peak) 8-core 8 4.2 550 40,773 23,224 75.5% Sun Fire X8420 SPEC OMPM2001 (peak) 16-core 16 4.7 570 94,350 25,932 263% HP AlphaServer GS1280 SPEC OMPM2001 (peak) Overall 64 5 595 242,116 104,714 88.8% Sun/Fujitsu M8000 SPEC OMPL2001 base (64-core) 64 2.30 p5-595 1,005,583 532,576 98.1% Sun/Fujitsu M8000 LINPACK HPC 2-core 2 1.90 p5-520 14.31 12.05 18.8% HP rx1620 (1.6 GHz) LINPACK HPC 4-core 4 4.7 520 65 21.71 199.4% HP rx5670 LINPACK HPC 8-core 8 5 550 137.6 48.55 183.4% HP rx6600 LINPACK HPC 16-core 16 5 570 277.7 88.8 212.7% HP rx8620 LINPACK HPC 32-core 32 4.7 575 500 268.6 86.1% Fujitsu/Sun M9000 LINPACK HPC 64-core 64 5 595 1050 342 207% HP Superdome Benchmarks Second Place System (non-IBM) POWER vs. Best Competitive Result Comparing the best available results vs. POWER Sources: http://www.spec.org http://www.tpc.org http://www.sap.com/benchmark/ http://performance.netlib.org/performance/html/PDSreports.html All results are as of 05/01/09 TPC-C results with processor chip/core/thread. SPEComp results: IBM cores = 2x chip, threads = 4x chip. SAP certification numbers can be found in SAP section of charts. Linpack results are SMP only. 46 © 2009 IBM Corporation