WHITE PAPER FUJITSU PRIMERGY SERVERS PERFORMANCE REPORT PRIMERGY RX600 S6

Transcription

1 WHITE PAPER PERFORMANCE REPORT PRIMERGY RX600 S6 WHITE PAPER FUJITSU PRIMERGY SERVERS PERFORMANCE REPORT PRIMERGY RX600 S6 This document contains a summary of the benchmarks executed for the PRIMERGY RX600 S6. The PRIMERGY RX600 S6 performance data are compared with the data of other PRIMERGY models and discussed. In addition to the benchmark results, an explanation has been included for each benchmark and for the benchmark environment. Version Contents Document history... 2 Technical data... 3 SPECcpu SPECjbb OLTP SAP SD vservcon VMmark V STREAM LINPACK Literature Contact Fujitsu Technology Solutions 2011 Page 1 (40)

2 Document history Version 1.0 First report version including the benchmark chapters SPECcpu2006 Measurements with Xeon E7-2803, E7-4807, E7-4820, E7-4830, E7-8837, E7-2850, E7-4850, E and E SPECjbb2005 Measurement with Xeon E vservcon Results for Xeon E7-8800, E and E processor series VMmark V2 Measurement with Xeon E Version 1.1 New benchmark chapters: OLTP-2 Results for E7-2803, E7-4807, E7-4820, E7-4830, E7-8837, E7-2850, E7-4850, E and E SAP SD Certification number STREAM Measurements with Xeon E7-2803, E7-2850, E7-4807, E7-4820, E7-4830, E7-4850, E7-4860, E and E LINPACK Measurements with Xeon E7-2803, E7-2850, E7-4807, E7-4820, E7-4830, E7-8837, E7-4850, E and E Updated benchmark chapters: SPECcpu2006 Additional measurements with Xeon E7-4807, E7-4820, E7-4830, E7-8837, E and E Version 1.2 Updated benchmark chapters: VMmark V2 New measurement with Xeon E Page 2 (40) Fujitsu Technology Solutions 2011

3 Technical data The PRIMERGY RX600 S6 is a quad-socket rack server with 4 height units that replaces the PRIMERGY RX600 S5. It has an Intel 7500 chip set, two to four Intel Xeon Series E or E processors or two Intel Xeon Series E processors, two to eight memory boards (each with eight DIMM slots) for up to 1 TB DDR3-SDRAM, two onboard 2-port 1-GBit Ethernet controllers, 10 PCI slots (three PCI-Express 2.0 x4, four PCI-Express 2.0 x8, one PCI-Express 2.0 x16 and two PCI-Express x4) and space for up to eight internal 2.5 drives (SSD or SAS HDD) or PCIe SSDs. Detailed technical information is available in the data sheet PRIMERGY RX600 S6. Fujitsu Technology Solutions 2011 Page 3 (40)

4 SPECcpu2006 Benchmark description SPECcpu2006 is a benchmark which measures the system efficiency with integer and floating-point operations. It consists of an integer test suite (SPECint2006) containing 12 applications and a floating-point test suite (SPECfp2006) containing 17 applications. Both test suites are extremely computing-intensive and concentrate on the CPU and the memory. Other components, such as Disk I/O and network, are not measured by this benchmark. SPECcpu2006 is not tied to a special operating system. The benchmark is available as source code and is compiled before the actual measurement. The used compiler version and their optimization settings also affect the measurement result. SPECcpu2006 contains two different performance measurement methods: the first method (SPECint2006 or SPECfp2006) determines the time which is required to process single task. The second method (SPECint_rate2006 or SPECfp_rate2006) determines the throughput, i.e. the number of tasks that can be handled in parallel. Both methods are also divided into two measurement runs, "base" and "peak" which differ in the use of compiler optimization. When publishing the results the base values are always used; the peak values are optional. Benchmark Arithmetics Type Compiler optimization SPECint2006 integer peak aggressive SPECint_base2006 integer base conservative SPECint_rate2006 integer peak aggressive SPECint_rate_base2006 integer base conservative SPECfp2006 floating point peak aggressive SPECfp_base2006 floating point base conservative SPECfp_rate2006 floating point peak aggressive SPECfp_rate_base2006 floating point base conservative Measurement result Speed Throughput Speed Throughput Application single-threaded multi-threaded single-threaded multi-threaded The measurement results are the geometric average from normalized ratio values which have been determined for individual benchmarks. The geometric average - in contrast to the arithmetic average - means that there is a weighting in favour of the lower individual results. Normalized means that the measurement is how fast is the test system compared to a reference system. Value 1 was defined for the SPECint_base2006-, SPECint_rate_base2006, SPECfp_base2006 and SPECfp_rate_base2006 results of the reference system. For example, a SPECint_base2006 value of 2 means that the measuring system has handled this benchmark twice as fast as the reference system. A SPECfp_rate_base2006 value of 4 means that the measuring system has handled this benchmark some 4/[# base copies] times faster than the reference system. "# base copies specify how many parallel instances of the benchmark have been executed. Not every SPECcpu2006 measurement is submitted by us for publication at SPEC. This is why the SPEC web pages do not have every result. As we archive the log files for all measurements, we can prove the correct implementation of the measurements at any time. Page 4 (40) Fujitsu Technology Solutions 2011

5 Benchmark results The PRIMERGY RX600 S6 was measured with Xeon series E7 processors. The benchmark programs were compiled with Intel C++/Fortran Compiler or and run under SUSE Linux Enterprise Server 11 SP1 (64-bit). All results have been published at Processor Cores GHz L3 cache QPI speed TDP SPECint_base2006 SPECint chips 4 chips 2 chips 4 chips Xeon E MB 4.80 GT/s 105 Watt Xeon E MB 4.80 GT/s 95 Watt Xeon E MB 5.86 GT/s 105 Watt Xeon E MB 6.40 GT/s 105 Watt Xeon E MB 6.40 GT/s 130 Watt Xeon E MB 6.40 GT/s 130 Watt Xeon E MB 6.40 GT/s 130 Watt Xeon E MB 6.40 GT/s 130 Watt Xeon E MB 6.40 GT/s 130 Watt All measurements with Intel C++/Fortran Compiler Processor Cores GHz L3 cache QPI speed TDP SPECint_rate_base2006 SPECint_rate chips 4 chips 2 chips 4 chips Xeon E MB 4.80 GT/s 105 Watt 219 1) 237 1) Xeon E MB 4.80 GT/s 95 Watt ) ) Xeon E MB 5.86 GT/s 105 Watt ) ) Xeon E MB 6.40 GT/s 105 Watt ) ) Xeon E MB 6.40 GT/s 130 Watt ) ) Xeon E MB 6.40 GT/s 130 Watt 427 1) 459 1) Xeon E MB 6.40 GT/s 130 Watt ) ) Xeon E MB 6.40 GT/s 130 Watt ) ) Xeon E MB 6.40 GT/s 130 Watt ) Intel C++/Fortran Compiler Fujitsu Technology Solutions 2011 Page 5 (40)

6 Processor Cores GHz L3 cache QPI speed TDP SPECfp_base2006 SPECfp chips 4 chips 2 chips 4 chips Xeon E MB 4.80 GT/s 105 Watt Xeon E MB 4.80 GT/s 95 Watt Xeon E MB 5.86 GT/s 105 Watt Xeon E MB 6.40 GT/s 105 Watt Xeon E MB 6.40 GT/s 130 Watt Xeon E MB 6.40 GT/s 130 Watt Xeon E MB 6.40 GT/s 130 Watt Xeon E MB 6.40 GT/s 130 Watt Xeon E MB 6.40 GT/s 130 Watt All measurements with Intel C++/Fortran Compiler Processor Cores GHz L3 cache QPI speed TDP SPECfp_rate_base2006 SPECfp_rate chips 4 chips 2 chips 4 chips Xeon E MB 4.80 GT/s 105 Watt Xeon E MB 4.80 GT/s 95 Watt 188 1) ) 394 Xeon E MB 5.86 GT/s 105 Watt 265 1) ) 551 Xeon E MB 6.40 GT/s 105 Watt 278 1) ) 579 Xeon E MB 6.40 GT/s 130 Watt 296 1) ) 637 Xeon E MB 6.40 GT/s 130 Watt Xeon E MB 6.40 GT/s 130 Watt 307 1) ) 634 Xeon E MB 6.40 GT/s 130 Watt 349 1) ) 720 Xeon E MB 6.40 GT/s 130 Watt 363 1) ) 752 1) Intel C++/Fortran Compiler The throughput with four processors both with the integer as well as the floating-point test suite is twice as large as that with two processors. SPECcpu2006: integer performance PRIMERGY RX600 S6 (4 sockets vs. 2 sockets) SPECint_rate SPECint_rate_base x Xeon E x Xeon E Page 6 (40) Fujitsu Technology Solutions 2011

7 SPECcpu2006: floating-point performance PRIMERGY RX600 S6 (4 sockets vs. 2 sockets) SPECfp_rate2006 SPECfp_rate_base x Xeon E x Xeon E The following four diagrams illustrate the throughput of the PRIMERGY RX600 S6 in comparison to its predecessor, the PRIMERGY RX600 S5, in the respective most performant configuration. SPECcpu2006: integer performance PRIMERGY RX600 S6 vs. predecessor SPECint SPECint_base PRIMERGY RX600 S5 4 x Xeon X7542 PRIMERGY RX600 S5 4 x Xeon X7560 PRIMERGY RX600 S6 4 x Xeon E Fujitsu Technology Solutions 2011 Page 7 (40)

8 SPECcpu2006: integer performance PRIMERGY RX600 S6 vs. predecessor SPECint_rate SPECint_rate_base PRIMERGY RX600 S5 4 x Xeon X7560 PRIMERGY RX600 S6 4 x Xeon E SPECcpu2006: floating-point performance PRIMERGY RX600 S6 vs. predecessor SPECfp SPECfp_base PRIMERGY RX600 S5 4 x Xeon X7542 PRIMERGY RX600 S5 4 x Xeon X7560 PRIMERGY RX600 S6 4 x Xeon E Page 8 (40) Fujitsu Technology Solutions 2011

9 SPECcpu2006: floating-point performance PRIMERGY RX600 S6 vs. predecessor SPECfp_rate SPECfp_rate_base PRIMERGY RX600 S5 4 x Xeon X7560 PRIMERGY RX600 S6 4 x Xeon E Benchmark environment All SPECcpu2006 measurements were made on a PRIMERGY RX600 S6 with the following hardware and software configuration: Hardware Model CPU Number of CPUs Primary cache Secondary cache Other cache Software Operating System PRIMERGY RX600 S6 Xeon E7-2803, E7-4807, E7-4820, E7-4830, E7-8837, E7-2850, E7-4850, E7-4860, E chips: Xeon E7-2803, E7-4807: Xeon E7-4820, E7-4830, E7-8837: all others: 4 chips: Xeon E7-4807: Xeon E7-4820, E7-4830, E7-8837: all others: 32 kb instruction + 32 kb data on chip, per core 256 kb on chip, per core 12 cores, 6 cores per chip 16 cores, 8 cores per chip 20 cores, 10 cores per chip 24 cores, 6 cores per chip 32 cores, 8 cores per chip 40 cores, 10 cores per chip Xeon E7-2803, E7-4807, E7-4820: 18 MB (I+D) on chip, per chip Xeon E7-4830, E7-8837, E7-2850, E7-4850, E7-4860: 24 MB (I+D) on chip, per chip Xeon E7-4870: 30 MB (I+D) on chip, per chip SUSE Linux Enterprise Server 11 SP1 (64-bit) Compilers Intel C++/Fortran Compiler and Some components may not be available in all countries or sales regions. Fujitsu Technology Solutions 2011 Page 9 (40)

10 SPECjbb2005 Benchmark description SPECjbb2005 is a Java business benchmark that focuses on the performance of Java Server platforms. SPECjbb2005 is essentially a modernized SPECjbb2000. The main differences are: The transactions have become more complex in order to cover a greater functional scope. The working set of the benchmark has been enlarged to the extent that the total system load has increased. SPECjbb2000 allows only one active Java Virtual Machine instance (JVM) whereas SPECjbb2005 permits several instances, which in turn achieves greater closeness to reality, particularly with large systems. On the software side SPECjbb2005 primarily measures the performance of the JVM used with its just-in-time compiler as well as their thread and garbage collection implementation. Some aspects of the operating system used also play a role. As far as hardware is concerned, it measures the efficiency of the CPUs and caches, the memory subsystem and the scalability of shared memory systems (SMP). Disk and network I/O are irrelevant. SPECjbb2005 emulates a 3-tier client/server system that is typical for modern business process applications with the emphasis on the middle-tier system: Clients generate the load, consisting of driver threads, which on the basis of TPC-C benchmark generate OLTP accesses to a database without thinking times. The middle tier system implements the business processes and the updating of the database. The database takes on the data management and is emulated by Java objects that are in the memory. Transaction logging is implemented on an XML basis. The major advantage of this benchmark is that it includes all three tiers that run together on a single host. The performance of the middle-tier is measured. Large-scale hardware installations are thus avoided and direct comparisons between the SPECjbb2005 results from the various systems are possible. Client and database emulation are also written in Java. SPECjbb2005 only needs the operating system as well as a Java Virtual Machine with J2SE 5.0 features. The scaling unit is a warehouse with approx. 25 MB Java objects. Precisely one Java thread per warehouse executes the operations on these objects. The business operations are assumed by TPC-C: New Order Entry Payment Order Status Inquiry Delivery Stock Level Supervision Customer Report However, these are the only features SPECjbb2005 and TPC-C have in common. The results of the two benchmarks are not comparable. SPECjbb2005 has 2 performance metrics: bops (business operations per second) is the overall rate of all business operations performed per second. bops/jvm is the ratio of the first metrics and the number of active JVM instances. In comparisons of various SPECjbb2005 results, both metrics must be specified. The following rules, according to which a compliant benchmark run has to be performed, are the basis for these three metrics: A compliant benchmark run consists of a sequence of measuring points with an increasing number of warehouses (and thus of threads) with the number in each case being increased by one warehouse. The run is started at one warehouse up through 2*MaxWh, but not less than 8 warehouses. MaxWh is the number of warehouses with the highest rate per second the benchmark expects. Per default the benchmark equates MaxWh with the number of CPUs visible by the operating system. The metric bops is the arithmetic average of all measured operation rates with MaxWh warehouses up to 2*MaxWh warehouses. Page 10 (40) Fujitsu Technology Solutions 2011

11 Benchmark results In April 2011 the PRIMERGY RX600 S6 with four Xeon E processors and 512 GB of PC3-8500R DDR3-SDRAM memory was measured. The measurement was taken under Windows Server 2008 R2 Enterprise with SP1. As JVM eight instances of the Java HotSpot(TM) 64-Bit Server VM on Windows, version 1.6.0_25 from Oracle was used. The following result was obtained: SPECjbb2005 bops = SPECjbb2005 bops/jvm = The PRIMERGY RX600 S6 achieved the best result of all 4-socket servers. 1 The following graphics illustrate the throughput of the PRIMERGY RX600 S6 in comparison to its predecessor PRIMERGY RX600 S5, in the respective most performant configuration. SPECjbb2005 bops: PRIMERGY RX600 S6 vs. predecessor SPECjbb2005 bops: PRIMERGY RX600 S6 vs. predecessor Benchmark environment The SPECjbb2005 measurement was run on a PRIMERGY RX600 S6 with the following hardware and software configuration: Hardware Model PRIMERGY RX600 S6 Processor Xeon E Number of chips 4 chips, 40 cores, 10 cores per chip, 2 threads per core Primary Cache 32 kb instruction + 32 kb data on chip, per core Secondary Cache 256 kb (I+D) on chip, per core Tertiary Cache 30 MB (I+D) on chip, per chip Memory 64 x 8 GB PC3-8500R DDR3-SDRAM Software Operating System Windows Server 2008 R2 Enterprise + SP1 JVM Version Oracle Java HotSpot(TM) 64-Bit Server VM on Windows, version 1.6.0_25 Some components may not be available in all countries or sales regions. 1 The comparative values mentioned above reflect the status of May 4th, This comparison is based on the SPECjbb2005 results of 4-socket servers. For the latest SPECjbb2005 results, visit Fujitsu Technology Solutions 2011 Page 11 (40)

12 OLTP-2 Benchmark description OLTP stands for Online Transaction Processing. The OLTP-2 benchmark is based on the typical application scenario of a database solution. In OLTP-2 database access is simulated and the number of transactions achieved per second (tps) determined as the unit of measurement for the system. In contrast to benchmarks such as SPECint and TPC-E, which were standardized by independent bodies and for which adherence to the respective rules and regulations are monitored, OLTP-2 is an internal benchmark of Fujitsu. OLTP-2 is based on the well-known database benchmark TPC-E. OLTP-2 was designed in such a way that a wide range of configurations can be measured to present the scaling of a system with regard to the CPU and memory configuration. Even if the two benchmarks OLTP-2 and TPC-E simulate similar application scenarios using the same load profiles, the results cannot be compared or even treated as equal, as the two benchmarks use different methods to simulate user load. OLTP-2 values are typically similar to TPC-E values. A direct comparison, or even referring to the OLTP-2 result as TPC-E, is not permitted, especially because there is no priceperformance calculation. Further information can be found in the document Benchmark Overview OLTP-2. Benchmark results The OLTP-2 values for the PRIMERGY RX600 S6 were determined for the Intel Xeon E7-48xx, E7-28xx and E processor series with memory configurations of 256 GB, 512 GB and 1024 GB. These results are based on the operating system Microsoft Windows Server 2008 R2 Enterprise SP1 and the database SQL Server 2008 R2 Enterprise x64 Edition. The database performance depends to a great degree on the configuration options of a system with hard disks and their controllers. Throughputs of the dimension specified here can be achieved if the typically external disk subsystem is not a bottleneck. Further information about the system configuration can be found in the section Benchmark environment. The table provides an overview of the processors including their properties that have been released for the PRIMERGY RX600 S6. Processor Cores/ Chip HT TM Processor frequency L3 Cache QPI speed Memory timing TDP E GHz 30 MB 6.4 GT/s 1066 MHz 130 watt E GHz 24 MB 6.4 GT/s 1066 MHz 130 watt E GHz 24 MB 6.4 GT/s 1066 MHz 130 watt E GHz 24 MB 6.4 GT/s 1066 MHz 105 watt E GHz 18 MB 5.86 GT/s 978 MHz 105 watt E GHz 18 MB 4.8 GT/s 800 MHz 95 watt E GHz 24 MB 6.4 GT/s 1066 MHz 130 watt E GHz 24 MB 6.4 GT/s 1066 MHz 130 watt HT = Hyper-Threading, TM = Turbo mode, QPI = QuickPath Interconnect, GT = Giga transfer, TDP = Thermal Design Power As regards memory, the maximum configuration with 16 GB modules and one reduced configuration with 8 GB and 4 GB modules are considered. Here, the timing only depends on the processor type and not on the type or number of memory modules used. Further information about memory performance can be found in the White Paper Memory Performance of Xeon E / 4800 / 2800 (Westmere-EX) Based Systems. A guideline in the database environment for selecting main memory is that sufficient quantity is more important than the speed of the memory accesses. Page 12 (40) Fujitsu Technology Solutions 2011

13 The following diagram shows the OLTP-2 performance data of the PRIMERGY RX600 S6, which can be achieved with two and four processors of the Intel Xeon series and various memory configurations. PRIMERGY RX600 S6 : OLTP-2 [tps] bold, cursive number: measured result others: calculated result E Core E Core E Core E Core E Core E Core E Core E Core E Core 4CPUs-1024GB 4CPUs-512GB 4CPUs-256GB 2CPUs-512GB 2CPUs-256GB It becomes evident that due to the large number of released processors the PRIMERGY RX600 S6 covers a wide performance range. If you compare the OLTP-2 value of the processor (E7-4807) with the lowest performance (1402 tps) and the processor (E7-4870) with the highest performance (2825 tps), both at maximum memory configuration and in the configuration with four processors, the result is an increase in performance by the factor of 2. Based on the results achieved the processors can be divided into different groups: The E7-4870, E and E processors with their ten cores per CPU are to be found at the top end of the performance range, followed by the E7-4830, E processors with eight cores. And at the lower end is the E as a processor with only six cores without Turbo mode. All these processors have Hyper- Threading. Database applications and the OLTP-2 load in particular benefit from doubling the logical processor cores through Hyper-Threading. An exception is the E processor, which has eight cores and a high clock frequency, but no Hyper-Threading, and almost achieves the performance of the E processor. A special position is held by the two processors E and E7-2803, which are only designed for the twofold and not for the four-fold configuration as well. The database throughput for all processor types could only be slightly increased by increasing the main memory. However, with a server with four sockets the question arises as to how good is the performance scaling from two to four processors. The better the scaling, the lower the overhead usually caused by the shared use of resources within a server. The scaling factor also depends on the application. If the server is used as a database server, an increase in performance of about 80% can be observed if you double the CPU number from two to four processors. Fujitsu Technology Solutions 2011 Page 13 (40)

14 If you compare the PRIMERGY RX600 S6 with its predecessors, the PRIMERGY RX600 S3, PRIMERGY RX600 S4 and PRIMERGY RX600 S5 with each at maximum configuration, the result is an increase in throughput of 747% (RX600 S3), 283% (RX600 S4) and 38% (RX600 S5) respectively OLTP-2: PRIMERGY RX600 S3 vs. RX600 S4 vs. RX600 S5 vs.rx600 S % % % PRIMERGY RX600 S3 4 Xeon 7140M 64 GB RAM PRIMERGY RX600 S4 4 Xeon GB RAM PRIMERGY RX600 S5 4 Xeon X GB RAM PRIMERGY RX600 S6 4 E GB RAM Page 14 (40) Fujitsu Technology Solutions 2011

15 Benchmark environment A typical OLTP-2 benchmark environment is shown symbolically in the following diagram: Driver Tier A Tier B Network Network Storage Subsystem Application Server Database Server Clients System under Test (SUT) Database Server (Tier B) Hardware System PRIMERGY RX600 S6 Processor 4 E (10C, 2.4 GHz) 4 E (10C, 2.27 GHz) 4 E (10C, 2.0 GHz) 4 E (8C, 2.13 GHz) 4 E (8C, 2.0 GHz) Memory Settings (default) Network interface 4 onboard LAN 1 Gb/s 4 E (6C, 1.87 GHz) 4 E (8C, 2.67 GHz) 2 E (10C, 2.0 GHz) 2 E (6C, 1.73 GHz) 1024 GB, 512 GB and 256 GB, registered ECC DDR3 (16 GB, 8 GB and 4 GB DIMMs) Turbo Mode enabled, NUMA Support enabled, Hyper-Threading enabled Disk subsystem RX600 S6: Onboard RAID Ctrl SAS 6G 5/6 512MB GB 15k rpm SAS Drive, RAID1 (OS) GB 10k rpm SAS Drive, RAID10 (LOG) 10 LSI MegaRAID SAS e 10 JX40: 24 64GB SSD Drive each, RAID5 (data) Software Operating system Windows Server 2008 R2 Enterprise SP1 Database SQL Server 2008 R2 Enterprise x64 Application Server (Tier A) Hardware System Processor Memory Network interface Disk subsystem Software Operating system 2 PRIMERGY RX200 S6 2 Xeon E5647 (4C, 2.93 GHz) 12 GB, 1333 MHz registered ECC DDR3 2 onboard LAN 1 Gb/s, 2 Dual Port LAN 1Gb/s 1 73 GB 15k rpm SAS Drive Windows Server 2008 R2 Standard Fujitsu Technology Solutions 2011 Page 15 (40)

16 Clients Hardware System 2 PRIMERGY RX200 S5 Processor 2 Xeon X5570 (4C, 2.93 GHz) Memory 24 GB, 1333 MHz registered ECC DDR3 Network interface 2 onboard LAN 1 Gb/s Disk subsystem 1 73 GB 15k rpm SAS Drive Software Operating system Windows Server 2008 R2 Standard OLTP-2 software EGen version Some components may not be available in all countries / sales regions. Page 16 (40) Fujitsu Technology Solutions 2011

17 SAP SD Benchmark description The SAP application software consists of modules used to manage all standard business processes. These include modules for ERP (Enterprise Resource Planning), such as Assemble-to-Order (ATO), Financial Accounting (FI), Human Resources (HR), Materials Management (MM), Production Planning (PP) plus Sales and Distribution (SD), as well as modules for SCM (Supply Chain Management), Retail, Banking, Utilities, BI (Business Intelligence), CRM (Customer Relation Management) or PLM (Product Lifecycle Management). The application software is always based on a database so that a SAP configuration consists of the hardware, the software components operating system, the database and the SAP software itself. SAP AG has developed SAP Standard Application Benchmarks in order to verify the performance, stability and scaling of a SAP application system. The benchmarks, of which SD Benchmark is the most commonly used and most important, analyze the performance of the entire system and thus measure the quality of the integrated individual components. The benchmark differentiates between a 2-tier and a 3-tier configuration. The 2-tier configuration has the SAP application and database installed on one server. With a 3-tier configuration the individual components of the SAP application can be distributed via several servers and an additional server handles the database. The entire specification of the benchmark developed by SAP AG, Walldorf, Germany can be found at: Fujitsu Technology Solutions 2011 Page 17 (40)

18 Benchmark results The certification number from SAP specifies that the PRIMERGY RX600 S6, equipped with 4 Xeon E processors, achieved the following result with SAP Enhancement Package 4 for SAP ERP 6.0 and SQL Server 2008 under Windows Server 2008 R2 Datacenter on 15 July 2011: Certification number Number of SAP SD benchmark users 13,285 Average dialog response time Throughput Fully processed order line items/hour dialog steps/hour SAPS 0.97 seconds 1,453,000 4,359,000 72,650 Average database request time (dialog/update) sec / sec CPU utilization of central server 99% Operating system, central server RDBMS SQL Server 2008 Windows Server 2008 R2 Datacenter SAP Business Suite software SAP enhancement package 4 for SAP ERP 6.0 Configuration Central server PRIMERGY RX600 S6 4 processors / 40 cores / 80 threads Xeon E GB main memory The following diagram illustrates the throughput of the PRIMERGY RX600 S6 in comparison to its predecessor, the PRIMERGY RX600 S5, in the respective most performant configuration. SAP SD: PRIMERGY RX600 S6 vs. predecessor PRIMERGY RX600 S6 4 x Xeon E GB RAM PRIMERGY RX600 S5 4 x Xeon X GB RAM Number of Benchmark Users Page 18 (40) Fujitsu Technology Solutions 2011

19 Benchmark environment 2-tier environment Load generator System Under Test System Under Test (SUT) Hardware Server PRIMERGY RX600 S6 Processor 4 Xeon E Memory Disk Subsystem Software Operating System Database 64 4 GB PC R DDR3-SDRAM 1 PRIMERGY RX600 S6: 2 RAID Ctrl SAS 6G 5/6 512MB (D2616) 2 RAID Contr BBU Upgrade for RAID 5/6 V70 5 HD SAS 6G 73GB 15K HOT PL 2.5" EP 1 HD SAS 6G 146GB 15K HOT PL 2.5" EP 1 HD SAS 6G 300GB 15K HOT PL 2.5" EP 2 FC Ctrl 8Gb/s 1 Chan LPe1250 MMF LC 1 FibreCAT CX4-480 Storage Unit Windows Server 2008 R2 Datacenter + SP1 SQL Server 2008 Enterprise Edition SAP Business Suite software SAP enhancement package 4 for SAP ERP 6.0 Load generator Hardware Model Processor Memory Software PRIMERGY RX300 S4 Operating system Linux x Xeon X5460, 3.17 GHz, 12 MB L2 cache 12 GB PC2-5300F DDR2-SDRAM Some components may not be available in all countries or sales regions. Fujitsu Technology Solutions 2011 Page 19 (40)

20 vservcon Benchmark description vservcon is a benchmark used by Fujitsu Technology Solutions to compare server configurations with hypervisor with regard to their suitability for server consolidation. This allows both the comparison of systems, processors and I/O technologies as well as the comparison of hypervisors, virtualization forms and additional drivers for virtual machines. vservcon is not a new benchmark in the true sense of the word. It is more a framework that combines already established benchmarks (or in modified form) as workloads in order to reproduce the load of a consolidated and virtualized server environment. Three proven benchmarks are used which cover the application scenarios database, application server and web server. Application scenario Benchmark No. of logical CPU cores Memory Database Sysbench (adapted) GB Java application server SPECjbb (adapted, with 50% - 60% load) 2 2 GB Web server WebBench GB Each of the three application scenarios is allocated to a dedicated virtual machine (VM). Add to these a fourth machine, the so-called idle VM. These four VMs make up a "tile". Depending on the performance capability of the underlying server hardware, you may as part of a measurement also have to start several identical tiles in parallel in order to achieve a maximum performance score. System Under Test Database VM Java VM Web VM Idle VM Tile n Database VM Database VM Database VM Java VM Java VM Java VM Web VM Web VM Web VM Idle VM Idle VM Idle VM Tile 3 Tile 2 Tile 1 Each of the three vservcon application scenarios provides a specific benchmark result in the form of application-specific transaction rates for the respective VM. In order to derive a normalized score, the individual benchmark results for one tile are put in relation to the respective results of a reference system. The resulting relative performance values are then suitably weighted and finally added up for all VMs and tiles. The outcome is a score for this tile number. Starting as a rule with one tile, this procedure is performed for an increasing number of tiles until no further significant increase in this vservcon score occurs. The final vservcon score is then the maximum of the vservcon scores for all tile numbers. This score thus reflects the maximum total throughput that can be achieved by running the mix defined in vservcon that consists of numerous VMs up to the possible full utilization of CPU resources. This is why the measurement environment for vservcon measurements is designed in such a way that only the CPU is the limiting factor and that no limitations occur as a result of other resources. The progression of the vservcon scores for the tile numbers provides useful information about the scaling behavior of the "System under Test". Moreover, vservcon also documents the total CPU load of the host (VMs and all other CPU activities) and, if possible, electrical power consumption. A detailed description of vservcon is in the document: Benchmark Overview vservcon. Page 20 (40) Fujitsu Technology Solutions 2011

21 E Series E Series 2800 WHITE PAPER PERFORMANCE REPORT PRIMERGY RX600 S6 VERSION: Benchmark results The current generation of PRIMERGY four and eight-socket systems is based on Intel Xeon series E7-2800, E and E processors. The configuration options of these systems vary, as can be seen in the following table. Processor RX600 S6 RX900 S2 6 Cores, HT E max Cores, HT, TM E max. 2 6 Cores, HT E max. 4 8 Cores, HT, TM 10 Cores, HT, TM E max. 4 E max. 4 E max. 4 E max. 4 E max. 4 8 Cores TM E max. 4 max. 8 HT, TM E max. 8 E max Cores, HT, TM E max. 8 E max. 8 HT = Hyper-Threading, TM = Turbo Mode The current generation of PRIMERGY four and eight-socket systems is very suitable for application virtualization thanks to the progress made in processor technology. Compared with a system based on the previous processor generation an approximate 22% higher virtualization performance can be achieved (measured in vservcon score) as 10-Core processors are also available. On the basis of the previously described vservcon profile almost optimal utilization of the CPU system resources is possible with 176 real application VMs (equivalent to 44 tiles) if the system is fully assembled with eight such processors. Fujitsu Technology Solutions 2011 Page 21 (40)

22 2 E E E E E E E E E7-2850/E E7-8867L 2 E E E E E7-8867L 4 E E E E E Final vservcon Score 8 E E WHITE PAPER PERFORMANCE REPORT PRIMERGY RX600 S6 VERSION: The first diagram compares the virtualization performance values that can be achieved with the individual processors. A large selection of released processors of the systems discussed here with six, eight or ten cores was considered #Tiles The relatively large performance differences between the processors as seen in the diagram can be explained by their features and the number of CPUs used. In the groups with the same number of CPUs the values scale on the basis of the number of cores, the size of the L3 cache and the CPU clock frequency. Furthermore, the data transfer rate between processors ("QPI Speed") also determines performance. The Xeon E processor shows itself to be weaker in the comparison when it comes to performance, because it has to do without Hyper-Threading (HT). It became evident that an increase in performance of almost 100% can be achieved by doubling the number of CPUs. More information about the topic "Memory Performance" and QPI architecture can be found in the White Paper Memory Performance of Xeon E / 4800 / 2800 (Westmere-EX) Based Systems. A guideline in the virtualization environment for selecting main memory is that sufficient quantity is more important than the speed of the memory accesses. The technical data of the processors is set out again clearly and concisely in the table below. Page 22 (40) Fujitsu Technology Solutions 2011

23 2800 E Series E Series WHITE PAPER PERFORMANCE REPORT PRIMERGY RX600 S6 VERSION: Processor #Cores/ L3 Cache Processorfrequency Chip QPI Speed HT TM TDP #Tiles Score 6 Cores, HT E7-2803(2 ) 6 18 MB 1.73 GHz 4.8 GT/s 105 W Cores, HT, TM E7-2850(2 ) MB 2.00 GHz 6.4 GT/s 130 W Cores, HT 8 Cores, HT, TM 10 Cores, HT, TM E7-4807(2 ) 6 18 MB 1.87 GHz 4.8 GT/s 95 W E7-4807(4 ) 6 18 MB 1.87 GHz 4.8 GT/s 95 W E7-4820(2 ) 8 18 MB 2.00 GHz 5.86 GT/s 105 W E7-4820(4 ) 8 18 MB 2.00 GHz 5.86 GT/s 105 W E7-4830(2 ) 8 24 MB 2.13 GHz 6.4 GT/s 105 W E7-4830(4 ) 8 24 MB 2.13 GHz 6.4 GT/s 105 W E7-4850(2 ) MB 2.00 GHz 6.4 GT/s 130 W E7-4850(4 ) MB 2.00 GHz 6.4 GT/s 130 W E7-4860(2 ) MB 2.27 GHz 6.4 GT/s 130 W E7-4860(4 ) MB 2.27 GHz 6.4 GT/s 130 W E7-4870(2 ) MB 2.40 GHz 6.4 GT/s 130 W E7-4870(4 ) MB 2.40 GHz 6.4 GT/s 130 W Cores TM HT, TM E7-8837(2 ) 8 24 MB 2.67 GHz 6.4 GT/s 130 W E7-8837(4 ) 8 24 MB 2.67 GHz 6.4 GT/s 130 W E7-8837(8 ) 8 24 MB 2.67 GHz 6.4 GT/s 130 W E7-8830(4 ) 8 24 MB 2.13 GHz 6.4 GT/s 105 W E7-8830(8 ) 8 24 MB 2.13 GHz 6.4 GT/s 105 W Cores, HT, TM E7-8850(4 ) MB 2.00 GHz 6.4 GT/s 130 W E7-8850(8 ) MB 2.00 GHz 6.4 GT/s 130 W E7-8860(4 ) MB 2.27 GHz 6.4 GT/s 130 W E7-8860(8 ) MB 2.27 GHz 6.4 GT/s 130 W E7-8870(4 ) MB 2.40 GHz 6.4 GT/s 130 W E7-8870(8 ) MB 2.40 GHz 6.4 GT/s 130 W QPI = QuickPath Interconnect, GT = Gigatransfer, HT = Hyper-Threading, TM = Turbo Mode, TDP = Thermal Design Power Fujitsu Technology Solutions 2011 Page 23 (40)

24 vservcon score vservcon score WHITE PAPER PERFORMANCE REPORT PRIMERGY RX600 S6 VERSION: The next two diagrams illustrate the virtualization performance for increasing numbers of VMs based on the PRIMERGY RX600 S6 Xeon E (8-core) and E (10-core) processors. The respective CPU loads of the host have also been entered. The number of tiles with optimal CPU load is typically at about 90%; beyond that you have overload, which is where virtualization performance no longer increases, or sinks again. Scores and CPU utilization x E vservconv score (left axis) CPU Util (right axis) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% # Tiles Scores and CPU utilization x E vservconv score (left axis) CPU Util (right axis) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% # Tiles In addition to the increased number of physical cores, Hyper-Threading, which is supported by almost all Xeon E7 processors (except E7-8837), is an additional reason for the high number of VMs that can be operated. As is known, a physical processor core is consequently divided into two logical cores so that the number of cores available for the hypervisor is doubled. This standard feature thus generally increases the virtualization performance of a system. The scaling curves for the number of tiles as seen in the previous diagram are specifically for systems with Hyper-Threading. 40 physical and thus 80 logical cores are available with the Xeon E processor; approximately four of them are used per tile (see Benchmark description). This means that a parallel use of the same physical cores by several VMs is avoided up to a maximum of about 10 tiles. That is why the performance curve in this range scales almost ideal. For the quantities above the growth is flatter up to CPU full utilization. Page 24 (40) Fujitsu Technology Solutions 2011

25 The previous diagram examined the total performance of all application VMs of a host. However, studying the performance from an individual application VM viewpoint is also interesting. This information is in the previous diagram. For example, the total optimum is reached in the above Xeon E situation with 72 application VMs (24 tiles, not including the idle VMs); the low load case is represented by three application VMs (one tile, not including the idle VM). Remember: the vservcon score for one tile is an average value across the three application scenarios in vservcon. This average performance of one tile drops when changing from the low load case to the total optimum of the vservcon score - from 2.12 to 23.90/24=0.99, i.e. to 47%. The individual types of application VMs can react very differently in the high load situation. It is thus clear that in a specific situation the performance requirements of an individual application must be balanced against the overall requirements regarding the numbers of VMs on a virtualization host. Fujitsu Technology Solutions 2011 Page 25 (40)

26 6 Tiles 18 Tiles 19 Tiles 24 Tiles vservcon Score WHITE PAPER PERFORMANCE REPORT PRIMERGY RX600 S6 VERSION: The virtualization-relevant progress in processor technology since 2008 has an effect on the one hand on an individual VM and, on the other hand, on the possible maximum number of VMs up to CPU full utilization. The following comparison shows the proportions for both types of improvements. Four selected configurations are compared: a PRIMERGY RX600 S4 from 2008 with 4 Xeon X7460, a PRIMERGY RX600 S5 from 2010 with 4 Xeon X7550 and a PRIMERGY RX600 S6 from 2011 with 4 Xeon E and 4 Xeon E respectively. Virtualization relevant improvements for 4-way servers 25 Few VMs (1 Tile) Score at optimum Tile count 20 x 2.97 x x X x GHz 2010 X x GHz 2011 E x GHz 2011 E x GHz 2008 X x GHz 2010 X x GHz 2011 E x GHz 2011 E x GHz Year CPU #Cores Freq. The progress for a few VMs (1 tile) is clear if one compares the 2008 system and the current system with the Xeon E processor. Despite 25% lower processor frequency (2.0 GHz instead of 2.67 GHz) the more recent system has a slightly higher vservcon score. One main reason for this is the new processor feature Extended Page Tables (EPT) 2. With a full load of the systems with VMs, the virtualization performance of the current system with Xeon E core processors is almost more than twice as high despite the 25% lower processor frequency. One cause is the performance increase that can be realized for an individual VM (see score for a few VMs). The other reason is that more than three times as many VMs are possible with total optimum (e. g. via Hyper-Threading and the multiplied memory bandwidth). However, the optimum is achieved with 19 in contrast to six tiles with a lower individual VM performance. The score increase of the current system compared to the system with the previous processor generation at full load can only be completely appreciated when the corresponding reduction in electrical power consumption is also taken into consideration. The guideline is usually seen in TDP (Thermal Design Power). The Xeon X7550 processor used in 2010 has TDP of 130 W, whereas TDP for the Xeon E processor currently in use is only 105 W for a comparable virtualization performance. 2 EPT accelerates memory virtualization via hardware support for the mapping between host and guest memory addresses. Page 26 (40) Fujitsu Technology Solutions 2011

27 vservcon score WHITE PAPER PERFORMANCE REPORT PRIMERGY RX600 S6 VERSION: A further increase in the virtualization performance can be obtained by using a processor with a higher number of cores. If one considers (also with full load) the same measurements with the current system for the 10-core processor instead of the 8-core processor, you can see a clear improvement in comparison to the 2008 system. We must explicitly point out that the increased virtualization performance as seen in the score cannot be completely deemed as an improvement for one individual VM. Performance increases in the virtualization environment since 2010 are mainly achieved by increases in the maximum number of VMs that can be operated. With a server with eight sockets the question arises as to how good is the performance scaling from four to eight processors. The better the scaling, the lower the overhead usually caused by the shared use of resources within a server. The scaling factor also depends on the application. If the server is used as a virtualization platform for server consolidation, scaling can be increased to a factor of When operating with eight processors, the PRIMERGY RX900 S2 thus reaches almost twice the performance level as with four processors, as seen in the following diagram based on the processor version Xeon E RX600 S6 4 E RX900 S2 4 E RX900 S2 8 E And in comparison to the PRIMERGY RX600 S6, which is already fully configured with four processors, the PRIMERGY RX900 S2 shows very good performance. The performance increase of the PRIMERGY RX900 S2 with eight processors in comparison to the PRIMERGY RX600 S6 with four processors is 83.68% (1.84 factor). The slight performance increase of the PRIMERGY RX600 S6 fully configured with four processors in comparison to the PRIMERGY RX900 S2 with four processors (1.93 factor) can be explained by the fact that the first system is designed for four processors, whereas a system, such as the PRIMERGY RX900 S2, has to scale efficiently up to eight processors. Fujitsu Technology Solutions 2011 Page 27 (40)

28 Benchmark environment The measurements were made with the environment described below: Framework controller Server Storage System Multiple 1Gb or 10Gb networks System under Test (SUT) Load generators All the vservcon scores for the Intel Xeon E7-8800, E and E processor series were determined by way of example on a PRIMERGY RX600 S6 and PRIMERGY RX900 S2. SUT hardware Model PRIMERGY RX600 S6 Processor 2 Xeon E (6C, 1.73 GHz) 2 Xeon E (10C, 2.00 GHz) 2 Xeon E (6C, 1.87 GHz) 4 Xeon E (6C, 1.87 GHz) 2 Xeon E (8C, 2.00 GHz) 4 Xeon E (8C, 2.00 GHz) 2 Xeon E (8C, 2.13 GHz) 4 Xeon E (8C, 2.13 GHz) 2 Xeon E (10C, 2.00 GHz) 4 Xeon E (10C, 2.00 GHz) 2 Xeon E (10C, 2.27 GHz) 4 Xeon E (10C, 2.27 GHz) 2 Xeon E (10C, 2.40 GHz) 4 Xeon E (10C, 2.40 GHz) 2 Xeon E (8C, 2.67 GHz) 4 Xeon E (8C, 2.67 GHz) Memory Full configuration with 8 GB DIMMs: 2 CPUs: 256 GB, 4 CPUs: 512 GB Network interface 1 1-Gbit LAN for control 1 10-Gbit LAN for load Disk subsystem No internal hard disks were used, but ETERNUS DX80 storage systems. One 50 GB LUN per tile for the virtual disk files of the VMs. Each LUN is a RAID 0 array consisting of 2 Seagate ST SS disks (15 krpm). Storage connection Via FC controller Emulex LPe12002 SUT software Operating system Hypervisor VMware ESX Server Version Version 4.1 U1, Build BIOS Version Aptio 0.99G, deviations from default: Performance/Watt=Traditional SUT: virtualization-specific details ESX settings General details Default Described in the Benchmark Overview vservcon. Page 28 (40) Fujitsu Technology Solutions 2011

29 SUT-Hardware Model Processor Memory Network interface Disk subsystem Storage connection SUT-Software PRIMERGY RX900 S2 4 Xeon E (8C, 2.13 GHz) 8 Xeon E (8C, 2.13 GHz) 4 Xeon E (8C, 2.67 GHz) 8 Xeon E (8C, 2.67 GHz) 4 Xeon E (10C, 2.00 GHz) 8 Xeon E (10C, 2.00 GHz) 4 Xeon E (10C, 2.27 GHz) 8 Xeon E (10C, 2.27 GHz) 4 Xeon E (10C, 2.40 GHz) 8 Xeon E (10C, 2.40 GHz) Full configuration with 8 GB DIMMs: 4 CPUs: 512 GB, 8 CPUs: 1024 GB 1 1-Gbit LAN for control 1 10-Gbit LAN for load No internal hard disks were used, but ETERNUS DX80 storage systems. One 50 GB LUN per tile for the virtual disk files of the VMs. Each LUN is a RAID 0 array consisting of 2 Seagate ST SS disks (15 krpm). Über FC-Controller Emulex LPe12002 Operating system Hypervisor VMware ESX Server Version Version 4.1 U1, Build BIOS Version 00.17, default settings SUT: virtualisierungsspezifische Details ESX settings General details Default Described in the Benchmark Overview vservcon. Load generator hardware Model 17 PRIMERGY BX920 S1 server blades (PRIMERGY BX900 chassis) Processor 2 Xeon X5570, 2.93 GHz each Memory 12 GB Network interface 3 1 Gbit LAN each Operating system Windows Server 2008 R2 Enterprise with Hyper-V Load generator VMs (per tile 3 load generator VMs on various server blades) Processor 1 logical CPU Memory 512 MB Network interface 2 1 Gbit LAN each Operating system Windows Server 2003 Enterprise Some components may not be available in all countries or sales regions. Fujitsu Technology Solutions 2011 Page 29 (40)

30 VMmark V2 Benchmark description VMmark V2 is a benchmark developed by VMware to compare server configurations with hypervisor solutions from VMware regarding their suitability for server consolidation. In addition to the software for load generation, the benchmark consists of a defined load profile and binding regulations. The benchmark results can be submitted to VMware and are published on their Internet site after a successful review process. After the discontinuation of the proven benchmark VMmark V1 in October 2010, it has been succeeded by VMmark V2, which requires a cluster of at least two servers and covers data center functions, like Cloning and Deployment of virtual machines (VMs), Load Balancing, as well as the moving of VMs with vmotion and also Storage vmotion. VMmark V2 is not a new benchmark in the actual sense. It is in fact a framework that consolidates already established benchmarks, as workloads in order to simulate the load of a virtualized consolidated server environment. Three proven benchmarks, which cover the application scenarios mail server, Web 2.0, and e-commerce were integrated in VMmark V2. Application scenario Load tool # VMs Mail server LoadGen 1 Web 2.0 Olio client 2 E-commerce DVD Store 2 client 4 Standby server (IdleVMTest) 1 Each of the three application scenarios is assigned to a total of seven dedicated virtual machines. Then add to these an eighth VM called the standby server. These eight VMs form a tile. Because of the performance capability of the underlying server hardware, it is usually necessary to have started several identical tiles in parallel as part of a measurement in order to achieve a maximum overall performance. A new feature of VMmark V2 is an infrastructure component, which is present once for every two hosts. It measures the efficiency levels of data center consolidation through VM Cloning and Deployment, vmotion and Storage vmotion. The Load Balancing capacity of the data center is also used (DRS, Distributed Resource Scheduler). The result of VMmark V2 is a number, known as a score, which provides information about the performance of the measured virtualization solution. The score reflects the maximum total consolidation benefit of all VMs for a server configuration with hypervisor and is used as a comparison criterion of various hardware platforms. This score is determined from the individual results of the VMs and an infrastructure result. Each of the five VMmark V2 application or front-end VMs provides a specific benchmark result in the form of applicationspecific transaction rates for each VM. In order to derive a normalized score the individual benchmark results for one tile are put in relation to the respective results of a reference system. The resulting dimensionless performance values are then averaged geometrically and finally added up for all VMs. This value is included in the overall score with a weighting of 80%. The infrastructure workload is only present in the benchmark once for every two hosts; it determines 20% of the result. The number of transactions per hour and the average duration in seconds respectively are determined for the score of the infrastructure workload components. In addition to the actual score, the number of VMmark V2 tiles is always specified with each VMmark V2 score. The result is thus as follows: Score@Number of Tiles, for example 4.20@5 tiles. A detailed description of VMmark V2 is available in the document Benchmark Overview VMmark V2. Page 30 (40) Fujitsu Technology Solutions 2011

31 tiles tiles tiles tiles tiles tiles tiles VMmark V2 Score WHITE PAPER PERFORMANCE REPORT PRIMERGY RX600 S6 VERSION: Benchmark results On November 29, 2011 Fujitsu achieved with a PRIMERGY RX600 S6 with Xeon E processors and VMware ESX 4.1 U1 a VMmark V2 score of 18.20@18 tiles in a system configuration with a total of 2 40 processor cores and when using two identical servers in the System under Test (SUT). Fujitsu thereby improved the result achieved on July 12, 2011 of 17.98@18 tiles on identical hardware. This score as well as the detailed results and configuration data can be seen at With this result the PRIMERGY RX600 S6 is from a VMmark V2 viewpoint the most powerful server in a matched pair configuration consisting of two identical hosts in the 80-core category (valid as of benchmark result publication date). The diagram shows the results of the PRIMERGY RX600 S6 in comparison 3 with the appropriate competitor systems. Compared with the Cisco result on comparable hardware, the increase is 1.11%. 2 x 40 Cores % Fujitsu PRIMERGY RX600 S6 4 Xeon E Cisco UCS C460 M2 4 Xeon E Fujitsu PRIMERGY RX600 S6 4 Xeon E HP ProLiant DL580 G7 4 Xeon E Cisco UCS C460 M2 4 Xeon E HP ProLiant DL580 G7 4 Xeon E Cisco UCS C460 M2 4 Xeon E The processors used, which with a good hypervisor setting could make optimal use of their processor features, were the essential prerequisites for achieving the PRIMERGY RX600 S6 result. These features include the extended page tables (EPT 4 ) and Hyper-Threading. All this has a particularly positive effect during virtualization. All VMs, their application data, the host operating system as well as additionally required data were on a powerful fibre channel disk subsystem from ETERNUS DX80 systems. If possible, the configuration of the disk subsystem takes the specific requirements of the benchmark into account. The use of SSDs (Solid State Disk) resulted in advantages in the number and response times of the hard disks used. The network connection of the load generators and the infrastructure workload connection between the hosts were implemented with the 10Gb LAN ports. All the components used were optimally attuned to each other. 3 The above comparisons for the competitor products reflect the status of 29th November The VMmark V2 results are available at 4 EPT accelerates memory virtualization via hardware support for the mapping between host and guest memory addresses. Fujitsu Technology Solutions 2011 Page 31 (40)

32 Benchmark environment The measurement set-up is symbolically illustrated below: Clients & Management Server(s) Storage System Multiple 1Gb or 10Gb networks Load Generators incl. Prime Client and Datacenter Management Server vmotion network System under Test (SUT) SUT hardware Number of servers 2 Model PRIMERGY RX600 S6 Processor 4 Xeon E (10-Core, 2.40 GHz) Memory 512 GB (64 8 GB per DIMM, 4R), 1066MHz registered ECC DDR3 Network interface 1 integrated Intel 82575EB quad port 1GbE adapter 1 Fujitsu D2755 dual port 10GbE adapter Disk subsystem 10 ETERNUS DX80 storage systems: a total of 220 hard disks (incl. SSDs) in several RAID-0 arrays. Storage connection 1 dual port FC controller Emulex LPe12002 SUT software Operating system Hypervisor VMware ESX Server ESX version VMware ESX v4.1 U1, Build BIOS version Rev. R0.99I / Rev. R Load generator hardware Model Processor Memory Network interface Operating system Details See disclosure 19 server blade PRIMERGY BX620 S4 2 Intel Xeon 5130, 2 GHz (except Prime Client with 2 Intel Xeon X5470, 3.33 GHz) 4 GB 1 1 GBit LAN each Microsoft Windows Server 2003 R2 Enterprise, updated with SP2 and KB Some components may not be available in all countries or sales regions. Page 32 (40) Fujitsu Technology Solutions 2011

33 STREAM Benchmark description STREAM is a synthetic benchmark that has been used for many years to determine memory throughput and which was developed by John McCalpin during his professorship at the University of Delaware. Today STREAM is supported at the University of Virginia, where the source code can be downloaded in either Fortran or C. STREAM continues to play an important role in the HPC environment in particular. It is for example an integral part of the HPC Challenge benchmark suite. The benchmark is designed in such a way that it can be used both on PCs and on server systems. The unit of measurement of the benchmark is GB/s, i.e. the number of gigabytes that can be read and written per second. STREAM measures the memory throughput for sequential accesses. These can generally be performed more efficiently than accesses that are randomly distributed on the memory, because the CPU caches are used for sequential access. Before execution the source code is adapted to the environment to be measured. Therefore, the size of the data area must be at least four times larger than the total of all CPU caches so that these have as little influence as possible on the result. The OpenMP program library is used to enable selected parts of the program to be executed in parallel during the runtime of the benchmark, consequently achieving optimal load distribution to the available processor cores. During implementation the defined data area, consisting of 8-byte elements, is successively copied to four types, and arithmetic calculations are also performed to some extent. Type Execution Bytes per step Floating-point calculation per step COPY a(i) = b(i) 16 0 SCALE a(i) = q b(i) 16 1 SUM a(i) = b(i) + c(i) 24 1 TRIAD a(i) = b(i) + q c(i) 24 2 The throughput is output in GB/s for each type of calculation. The differences between the various values are usually only minor on modern systems. In general, only the determined TRIAD value is used as a comparison. The measured results primarily depend on the clock frequency of the memory modules; the CPUs influence the arithmetic calculations. The accuracy of the results is approximately 5%. This chapter specifies throughputs on a basis of 10 (1 GB/s = 10 9 Byte/s). Fujitsu Technology Solutions 2011 Page 33 (40)

34 Benchmark results The PRIMERGY RX600 S6 was measured with processors from the Xeon E7 series. The benchmark was compiled using the Intel C compiler 12.0 and performed under SUSE Linux Enterprise Server 11 SP1 (64- bit). The data area consisted of 120 million elements, which is equivalent to about 900 MB. Processor Cores GHz L3 cache QPI Speed TDP Xeon E MB 4.80 GT/s 105 Watt 32.6 Xeon E MB 6.40 GT/s 130 Watt 46.2 TRIAD [GB/s] 2 chips 4 chips Xeon E MB 4.80 GT/s 95 Watt 65.0 Xeon E MB 5.86 GT/s 105 Watt 80.8 Xeon E MB 6.40 GT/s 105 Watt 86.9 Xeon E MB 6.40 GT/s 105 Watt 88.2 Xeon E MB 6.40 GT/s 130 Watt 94.6 Xeon E MB 6.40 GT/s 130 Watt 101 Xeon E MB 6.40 GT/s 130 Watt 102 The following diagram illustrates the throughput of the PRIMERGY RX600 S6 in comparison to its predecessor, the PRIMERGY RX600 S5, in its most performant configuration. GB/s PRIMERGY RX600 S5 4 Xeon X7560 PRIMERGY RX600 S6 4 Xeon E STREAM TRIAD Page 34 (40) Fujitsu Technology Solutions 2011

35 Benchmark environment All STREAM measurements were based on a PRIMERGY RX600 S6 with the following hardware and software configuration: Hardware Model CPU Number of cores Primary Cache Secondary Cache Other Cache Memory Software Operating system PRIMERGY RX600 S6 Xeon E7-2803, E7-2850, E7-4807, E7-4820, E7-4830, E7-8837, E7-4850, E7-4860, E chips: Xeon E7-2803: Xeon E7-2850: 4 chips: Xeon E7-4807: Xeon E7-4820, E7-4830, E7-8837: All others: 12 cores 20 cores 24 cores 32 cores 40 cores 32 kb instruction + 32 kb data on chip, per core 256 kb on chip, per core Xeon E7-2803, E7-4807, E7-4820: Xeon E7-4870: All others: 64 8 GB PC3-8500R DDR3-SDRAM SUSE Linux Enterprise Server 11 SP1 (64-bit) Compiler Intel C Compiler 12.0 Benchmark Stream.c Version 5.9 Some components may not be available in all countries or sales regions. 18 MB (I+D) on chip, per chip 30 MB (I+D) on chip, per chip 24 MB (I+D) on chip, per chip Fujitsu Technology Solutions 2011 Page 35 (40)

36 LINPACK Benchmark description LINPACK was developed in the 1970s by Jack Dongarra and some other people to show the performance of supercomputers. The benchmark consists of a collection of library functions for the analysis and solution of linear system of equations. A description can be found in the document LINPACK can be used to measure the speed of a computer during the solution of an N dimensional linear system of equations. The result is specified in GFlops (Giga Floating Point Operations per Second). It is a measure of how many floating-point operations can be carried out per second. The number of floating-point operations required for the solution is determined by the formula 2 / 3 N N 2. For the calculation LINPACK requires a matrix of size N N in the main memory with the value N standing for the number of equations to be solved. Maximum performance is achieved if the available main memory can be fully used as a result of choosing this value. However, the determination of this limit is very timeconsuming and the expected increase in the result is only minor. The memory bandwidth of the system also has hardly any impact on the result, because floating-point calculations are chiefly carried out during the run and data exchange only seldom takes place between the parallel processes. Thus the benchmark result is determined for a value of N that is somewhat below the maximum value. LINPACK is classed as one of the leading benchmarks in the field of high performance computing (HPC). LINPACK is one of the seven benchmarks currently included in the HPC Challenge benchmark suite, which takes other performance aspects in the HPC environment into account. An Intel-optimized LINPACK version for individual systems is used on PRIMERGY servers. This is part of the Intel Compiler or can also be directly downloaded from It is possible to publish LINPACK results at Prerequisite for this is the use of an MPIbased (Message Passing Interface) version. (See: The maximum theoretical performance of a processor core follows from the number of floating-point operations that are performed within a clock cycle. Thus e.g. a processor with a clocking frequency of 2.4 GHz and 4 floating-point operations per cycle would achieve a maximum performance of 9.6 GFlops. The ratio of the measured result to the maximum value shows the efficiency of the system for floating-point calculations. The fewer memory accesses required during the calculation, the better the ratio. Experience shows that current processor architectures achieve approximately 90%. Page 36 (40) Fujitsu Technology Solutions 2011

37 Benchmark results The PRIMERGY RX600 S6 was measured with processors from the Xeon E7 series. The benchmark comes from the Intel Compiler 12.0 package and was performed under SUSE Linux Enterprise Server 11 SP1 (64- bit). The measured processors with SSE4.2 technology achieve 4 floating-point calculations per clock cycle. Thus, the theoretically achievable value is: GFlops max = 4 number of processor cores CPU frequency in GHz. The available main memory of 512 GB permits a dimension of N = Processors Cores GHz L3 cache [MB] QPI speed TDP Theor. Max. [GFlops] LINPACK [GFlops] Efficiency [%] Xeon E GT/s 105 Watt Xeon E GT/s 130 Watt Processors Cores GHz L3 cache [MB] QPI speed TDP Theor. Max. [GFlops] LINPACK [GFlops] Efficiency [%] Xeon E GT/s 95 Watt Xeon E GT/s 105 Watt Xeon E GT/s 105 Watt Xeon E GT/s 130 Watt Xeon E GT/s 105 Watt Xeon E GT/s 130 Watt Xeon E GT/s 130 Watt The results show that all the processors achieve more than 90% of the theoretical value and that the PRIMERGY RX600 S6 performs very well in the field of floating-point calculations. The following diagram illustrates the throughput of the PRIMERGY RX600 S6 in comparison to its predecessor, the PRIMERGY RX600 S5, in its most performant configuration. LINPACK: PRIMERGY RX600S6 vs. predecessor GFlops PRIMERGY RX600 S5 4 Xeon X7560 PRIMERGY RX600 S6 4 Xeon E Fujitsu Technology Solutions 2011 Page 37 (40)