Fujitsu PRIMERGY TX150 Torre
Contents
1. Fujitsu Technology Solutions 2009 Server PRIMERGY TX150 S6 o Storage 2 x 2 channel FC controllers t QLE2462 2 x FibreCAT CX500 180 data disks FC RAID O Onboard LAN 1Gbit 36 GB 15 krpm 4 log disks SAS RAID 10 146 GB 10 krpm Page 32 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Literature PRIMERGY Systems http ts fujitsu com primergy PRIMERGY TX150 S6 http docs ts fujitsu com dl aspx id af1d8493 dfbb 478e 89c9 92e91 7929234 PRIMERGY Performance http ts fujitsu com products standard servers primergy bov html Benchmark Overview OLTP 2 http docs ts fujitsu com dl aspx id e6f a4c9 aff6 4598 b199 836053214d93f OLTP 2 http www spec org osg cpu2006 Benchmark Overview SPECcpu2006 http docs ts fujitsu com dl aspx id 1a427c16 12bf 41b0 9ca3 4cc360ef14ce SPECcpu2006 http www spec org jbb2005 Benchmark Overview SPECjbb2005 http docs ts fujitsu com dl aspx id 54 1 1e8f9 8c56 4ee9 9b3b 98981 ab3e820 SPECjbb2005 SPECpower_ssj2008 http www spec org power ssj2008 http www spec org web2005 Benchmark Overview SPECweb2005 http docs ts fujitsu com dl aspx id efbe8db4 7b1b 481e bdee 66bdfa624b57 SPECweb2005 Performance Report Modular RAID for PRIMERGY StorageBench http docs ts fujitsu com dl aspx id 8f6d5779 2405 4cdd 8268 11948ba050e6 http www iometer org Contact PRIMERGY Hardware PRIMERGY Product Marketing mailto2. The ratio of power consumption between the different CPUs changes with every additional 1096 of the target load During active idle the difference is very small That relates to the power management features of the CPUs and the operating system They enable the CPUs to scale down the frequency and core voltage to a level where the CPUs consume the lowest power provided that the CPUs are idle So the power consumption for the different CPUs is almost the same at active idle As you can see this is not true for the Xeon X3220 processor This is due to the fact that this processor is based on another manufacturing technology 65 nm compared to 45 nm on all other processors The 45 nm manufac turing technology enables the processors to consume less power and brings additional power management features At the higher load levels the influence of the power management features is only marginal This is exactly where the Xeon X3360 processor can play to its strength Although it has the same TDP of 95 watts as the Xeon X3220 and X3370 the Xeon X3360 processor consumes much lower power at higher load levels This is related to the lower frequency com pared to the Xeon X3370 processor and to the less power consuming manufacturing technology of 45 nm compared to 65 nm in the case of the Xeon X3230 processor At 100 load the power consumption difference to the Xeon X3370 processor is 23 watts When looking at the Xeon E3120 processor you can see that it consumes th
3. data throughput in megabytes per second in short MB s transaction rate in I O operations per second in short IO s and latency time or also mean access time in ms For sequential load profiles data throughput is the normal indicator whereas for random load profiles with their small block sizes the transaction rate is normally used Throughput and transaction rate are directly proportional to each other and can be calculated according to the formula Data throughput MB s Transaction rate Disk l O s x Block size MB Transaction rate Disk l O s j d Data throughput MB s Block size MB Fujitsu Technology Solutions 2009 Page 22 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Benchmark results The PRIMERGY TX150 S6 is equipped with controllers from the Modular RAID family The variety of the RAID solu tions enables the user to choose the right controller for his application scenario The PRIMERGY TX150 S6 has the following RAID solutions to offer 1 SATA RAID onboard controller The controller is implemented directly on the motherboard of the server in the Intel ICH9R chipset and the RAID stack is realized by the server CPU This RAID solution is only foreseen for the connection of SATA hard disks Support is provided for RAID levels 0 1 and 10 as well as for RAID 5 with an additional iButton This controller does not have a controller cache 2 RAID Controlle
4. 115 Celeron Pentium DC Pentium DC Pentium DC Core 2 Duo Core 2 Duo Core 2 Duo 440 E2140 E2160 E2200 E4500 E7200 E4500 1core 2 cores 2 cores 2 cores 2 cores 2 cores 2 cores 2 GHz 1 60 GHz 1 80 GHz 2 20 GHz 2 20 GHz 2 53 GHz 2 20 GHz 800 MHz FSB 800 MHz FSB 800 MHz FSB 800 MHz FSB 800 MHz FSB 1067 MHz FSB 800 MHz FSB Y 2MB L2 cache 1MB L2 cache 1MB L2 cache 1MB L2 cache 2 MB L2 cache 3 MB L2 cache 2 MB L2 cache 35 Watt 65 Watt 65 Watt 65 Watt 65 Watt 65 Watt 65 Watt 175 150 25 00 T5 50 25 OLTP 2 PRIM ERGY TX150 S6 Xeon Xeon Xeon Xeon Xeon Xeon Xeon Xeon 3065 E3 110 E3120 X3210 X3220 X3350 X3360 X3370 2 cores 2 cores 2 cores 4 cores 4 cores 4 cores 4 cores 4 cores 2 33 GHz 3 00 GHz 3 16 GHz 2 13 GHz 2 40 GHz 2 66 GHz 2 83 GHz 3 00 GHz 1333 MHz FSB 1333 MHz FSB 1333 MHz FSB 1067 MHz FSB 1067 MHz FSB 1333 M Hz FSB 1333 M Hz FSB 1333 M Hz FSB 4 MB L2 cache 6 MB L2 cache 6 MB L2 cache 8 MB L2 cache 8 MB L2 cache 12 MB L2 cache 12 MB L2 cache 12 MB L2 cache 65 Watt 65 Watt 65 Watt 95 Watt 95 Watt 95 Watt 95 Watt 95 Watt Fujitsu Technology Solutions 2009 Page 31 33 White Paper Performance Report PRIMERGY TX150 S6 Benchmark environment Version 5 1 November 2008 Microsoft Windows Server 2008 Enterprise x64 Edition and database SQL Server 2008 Enterprise x64 Edition Clients 2 x PRIMERGY Econel 200 with 2 x Xeon 3 40 GHz 2 MB L2 cache 2 GB RAM onboard LAN 1Gbit f h Switch d
5. Normalized means measuring how fast the test system runs in comparison to a reference system The value of 1 was determined for the SPECint base2006 SPECint rate base2006 SPECfp base2006 and SPECfp rate base2006 results of the reference system Thus a SPECint base2006 value of 2 means for example that the measuring system has executed this benchmark approximately twice as fast as the reference system A SPECfp rate base2006 value of 4 means that the measuring system has executed this benchmark about 4 base copies times as fast as the reference system base copies here specifies how many parallel instances of the bench mark have been executed We do not submit all SPECcpu2006 measurements for publication at SPEC So not all results appear on SPEC s web sites As we archive the log data for all measurements we are able to prove the correct realization of the measurements any time Benchmark results The PRIMERGY TX150 S6 was measured with eight different processor versions e Celeron 440 Conroe L 1 core per chip 72 MB L2 cache per chip e Pentium Dual Core E2140 E2160 and E2200 Allendale 2 cores per chip 1 MB L2 cache per chip e Core 2 Duo E4500 and E4600 Allendale 2 cores per chip 2 MB L2 cache per chip e Core 2 Duo E7200 Wolfdale 2 cores per chip 3 MB L2 cache per chip e Xeon 3065 Conroe 2 cores per chip 4 MB L2 cache per chip SPEC SPECint SPECfp and the SPEC logo are registered trademark
6. SOOO Dell PowerEdge Redo PRIMERGY T150 Sp Xeon Za Xeon x35 1v M BEA Jeck 6 0 F2T 4 0 2 vhs BEA JEBackit 6 0 F2T 4 0 Source http www spec org jbb2005 results as of February 14 2008 Competitive benchmark results stated above reflect results published as of Feb 14 2008 The comparison presented above is based on the best performing servers with one Quad Core processor currently shipping by Dell and Fujitsu Siemens Computers now operating under the name of Fujitsu For the latest SPECjbb2005 benchmark results visit http www spec org bb2005 results Fujitsu Technology Solutions 2009 Page 9 33 Version 5 1 November 2008 White Paper Performance Report PRIMERGY TX150 S6 In August 2008 the PRIMERGY TX150 S6 was measured with the Xeon X3370 processor and a memory of 8 GB PC2 6400 DDR2 SDRAM The measurement was taken under Windows Server 2003 R2 Enterprise x64 Edition As JVM two instances of JRockit R 6 0 P27 5 0 build P27 5 0 5 97156 1 6 0 03 20080403 1524 windows x86 64 by BEA were used SPECIbb2005 bops PRIMERGY TX150 56 vs PRIMERGY TX150 55 PRIMERGY Tx150 6 200000 Xeon Aari evils BEA Joch 6 0 PaT 5 u bopz 50000 PRIMERGY T2150 Sp Xeon X335 e dY Me BEA Joch 6 0 Pa2T 4 0 100000 PRIMERGY T x150 Sp Xeon 2210 2 vM BEA Joch 6 0 P2T 4 0 5 OUO PRIMERGT Tx150 35 eon SOTO 10v M BES JRockit 5 0 PaT 2 0 warehouzesz When comparing the PRIMERGY TX150 S6 and its predecessor the PRIMERGY TX15
7. GAS GE En Pu 256 Na Er SAS 1078 512 ME um Er CG d e Ee Ge T Optimal 125 103 32 22 i25 fM 34 234 i25 23 34 Optimal Cache Settings Far LS S45 1078 Write back WO direct Disk cache enabled Access Pattern a Streaming sequential 54 KB 10 read b Restore sequential 64 EB 100 write c Database random SEB 6fit read 334 write d File Server random 5S4 KB B read 335 write Optimal Cache Settings For LS 545 1068 Disk cache enabled the LSI MegaRAID SAS 1078 controller with a 256 MB controller cache is about 3 2 with optimal cache settings and sequential write access The difference in performance to the LSI MegaRAID SAS 1068 controller is approximately 2 2 If the LSI MegaRAID 1078 controller with 256 MB and 512 MB cache is compared with the LSI MegaRAID 1068 control ler during random access with 8KB and 64KB then the throughput difference is between 5 and 6 5 Fujitsu Technology Solutions 2009 Page 27 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 When the onboard SATA ICH9R controller is compared with the LSI MegaRAID SAS 1068 controller it is evident that both controllers offer roughly the same performance when measurements are carried out with the same SATA hard disks and optimum cache settings The variations in performance are within the range of measuring accuracy Onboard SATA ICHAR and LSI SAS 1068 Relevant differences in through
8. Primergy PM ts fujitsu com PRIMERGY Performance and Benchmarks PRIMERGY Performance and Benchmarks mailto primergy benchmark ts fujitsu com Delivery subject to availability specifications subject to change without Published by department Internet notice correction of errors and omissions excepted http ts fujitsu com primergy All conditions quoted TCs are recommended cost prices in EURO excl VAT Enterprise Products iv Extranet unless stated otherwise in the text All hardware and software names used PRIMERGY Server hitri isst aps ta full eomiceneredu cis same are brand names and or trademarks of their respective holders PRIMERGY Performance Lab are inrined mailto primergy benchmark ts fujitsu com ud Copyright Fujitsu Technology Solutions GmbH 2009
9. SPECjbb2005 measures the implementations of the JVM JIT Just In Time compiler garbage collection threads and some aspects of the operating system As far as hardware is concerned it measures the effi ciency 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 em phasis on the middle tier system e Clients generate the load consisting of driver threads which on the basis of the TPC C benchmark generate OLTP accesses to a database without thinking times e The middle tier system implements the business processes and the updating of the database e 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 perform ance of the middle tier is measured thus avoiding large scale hardware installations and making direct comparisons possible between SPECjbb2005 results of different systems 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 exe
10. disk with 7 2 krpom you can then see that the throughput of the 372 SAS hard disk is about 2 8 times higher for random ac cess with 8 KB blocks and enabled disk cache than with the SATA hard disk LSI MegaRAID SAS 1078 The RAID array defines the way in which data is treated as regards availability How quickly the data is transferred in the respective RAID array context depends largely on the data throughput of the hard disks The number of hard disks con figured for the measurements in a RAID array was defined depending on the RAID level Between two and four hard disks were used To ensure that the hard disks do not represent a bottleneck when determining the efficiency of the controller under various cache settings the measurements were performed with 277 hard disks with a rotational speed of 15 krpm Throughput can in part be considerably increased through the cache settings However these increases in throughput differ depending on the data structure and access pattern For the measurements the controller cache option Read Mode is always set to No Read ahead The options Write Mode I O cache and Disk cache were varied The following diagram shows the throughputs for sequential read and write with 64 KB blocks and for different cache settings in RAID 1 with two and in RAID 5 with four 272 SAS hard disks The read throughput in RAID 1 with optimum cache settings is in the range of the RAID 1 2 HDs and RAID 5 4
11. disks are put together to form a Redundant Array of Independent Disks known as RAID in short with the data being spread over several hard disks in such a way that all the data is retained even if one hard disk fails except with RAID 0 The most usual methods of organizing hard disks in arrays are the RAID levels RAID 0 RAID 1 RAID 5 RAID 6 RAID 10 RAID 50 and RAID 60 Information about the basics of various RAID arrays is to be found in the paper Performance Report Modular RAID for PRIMERGY Depending on the number of disks and the installed controller the possible RAID configurations are used for the StorageBench analyses of the PRIMERGY servers For systems with two hard disks we use RAID 1 and RAID 0 for three and more hard disks we also use RAID 1E and RAID 5 and where applicable further RAID levels provided that the controller supports these RAID levels Regardless of the size of the hard disk a measurement file with the size of 8 GB is always used for the measurement In the evaluation of the efficiency of UO subsystems processor performance and memory configuration do not play a significant role in today s systems a possible bottleneck usually affects the hard disks and the RAID controller and not CPU and memory Therefore various configuration alternatives with CPU and memory need not be analyzed under StorageBench Measurement results For each load profile StorageBench provides various key indicators e g
12. execution of the measurements at any time The adjoining diagram shows the result SPECpower_ssj2006 PRIMERGY TX150 36 graph of the configuration described above Et EE measured with the PRIMERGY TX150 S6 SCH ED 750 t 1250 1500 1750 The red horizontal bars show the perform ance to power ratio in ssj ops watt upper 4 121 overall en ops watt x axis for each target load level which are tagged on the y axis of the diagram The blue line shows the run of the curve for the average power consumption bottom x axis at each target load level marked with a small rhomb The diagram shows how the efficiency of the server decreases with each target load level from 100 to active idle in 10 seg ments The black vertical line shows the benchmark result of 1 124 overall ssj_ops watt for the PRIMERGY TX150 S6 This is calculated by adding the measured transaction throughputs for each segment and then dividing by the sum of the aver age power consumed for each segment Target Load 10 a0 30 4 50 6D it CA 30 100 110 Average Power W The configuration was tuned to get the best possible result for this server in terms of performance per watt The memory configuration with 2 x 2 GB was selected to meet the criteria of best performance at lowest power consumption by popu lating only one slot of each available memory channel This configuration enables the benchmark to use the full capacity of the available memory bandwidth and at
13. hard disks were connected with the onboard SATA controller The measurement was performed using the HTTP software Accoria Rock Web Server v1 4 6 x86 64 under Red Hat Enterprise Linux 5 1 2 6 18 53 el5 x86 64 In May 2008 the PRIMERGY TX150 S6 was measured with one Xeon X3360 processor and 8 GB PC2 6400 DDR2 SDRAM Two quad port Intel PRO 1000GT and one Broadcom NetXtreme Il BCM5708 onboard were used for the network Two FibreCAT SX88 with 24 hard disks which were connected via an Emulex LPe11002 fibre channel control ler were used as disk subsystem A RAID 0 was built across the 24 hard disks The logging was done on a hard disk of type Seagate ST380013AS The operating system was resident on a Seagate ST3160815AS hard disk Both hard disks were connected with the onboard SATA controller The measurement was performed using the HTTP software Accoria Rock Web Server v1 4 6 x86 64 under Red Hat Enterprise Linux 5 1 2 6 18 53 el5 x86 64 In the class of servers with one Quad Core processor the PRIMERGY TX150 S6 achieved the best result SPECweb 2005 PRIMERGY TX150 56 vs other servers with 1 Quad Core processor Fumjitzu Siemens Compaters PRIMERGY T2150 6 Xeon 43360 6 GB PC2 6400 DDRZ SDRAIM Fmjitzu Sicemenz Compaters PRIMERGT TX150 6 Aeon X322 amp GB PC2 6400 DDRZ SDRAM Dell PowerEdge G60 Xeon X322 6 GB PC2 A4200 DDR2 SDRAIM S100 10000 ood 20000 Source http www spec org web2005 results as of Aug
14. has a significant influence on power con sumption but other configuration details as well Doubling the memory for example from 2 x 2 GB to 4 x 2 GB increases the power consumption at all load levels by about 6 7 watts on an average Adding three additional hard disks to the configuration increases the power consumption by about 20 24 watts per load level on an average When running the benchmark with all available power management features disabled in the BIOS and in the OS the PRIMERGY TX150 S6 consumes up to 15 watts more in the range of active idle to 50 load The PRIMERGY TX150 S6 is available with two different types of PSUs The first is a single standard PSU and the second are redundant PSUs By replacing the stan dard PSU with one single redundant PSU the power consumption increases by up to 30 watts This is explained by the fact that redundant power supplies need additional electronics e g power backplane for handling the redundancy and hot plug functionality and the other important fact is that the redundant PSUs have a different efficiency compared to standard PSUs When enabling full redundancy by adding the second redundant PSU the PRIMERGY TX150 S6 con sumes another 25 30 watts more The reason is that the PSUs share the load i e each of them gets only half the load and thus runs in a lower load range at a lower efficiency The final energy efficiency diagram shows the performance to power ratio power efficiency of all the previousl
15. the increase in throughput is a little higher than in the case with random access with 64 KB blocks and is roughly 23 with single disk 36 with RAID 0 and about 20 with RAID 1 MBs mHDCacheoH 074 13 os 630 e25 SE EHDCacheon 09 154 037 857 760 663 mmm SEB random 57 reads SO RAID ARAO B4 KE d A f SO RAID RAID 1 The onboard SATA RAID ICH9R controller does not have a controller cache This fact becomes particularly evident during write access in RAID 5 The enabling of this disk cache only brings about a moderate increase in throughput of RAID 5 and RAID 10 with four Hard Disks sequential Access pee P B4 KB sequential GA KB sequential reads writes RAIDS RAID i a RAIDOS RAID TO DHOCashe otf T2449 M712 665 1040 mHDCacheon 124 22 WTi5 2053 137418 about three fold Measurements with the LSI MegaRAID SAS 1078 controller which has a controller cache have shown that as a result of the enabled controller cache throughput increases of up to 39 fold and more are possi ble On the other hand with RAID 10 which actually con sists of two striped RAID 1 arrays the enabling of the disk cache fully benefits write throughput which increases by about 13 fold In contrast the disk cache has no impact on the through puts for sequential read with 64 KB blocks In RAID 10 better throughputs are achieved than in RAI
16. 0 S5 both in their highest per formance configurations an increase of 16496 is noted In all measurements the overall benchmark result includes the measured values from 2 to 4 warehouses TQ SPECjbb2005 bops PRIMERGY TX150 56 vs PRIMERGY TX150 55 id TUE Bagong 121492 T0000 164 _ 50000 PRIMERGY T2150 26 Xeon vo PO PRIMERGY T2150 Sp Xcon Atel eon SOTO TIVM BEA JRockitS 0 23JNMhd BEA JAockit 6 0 2 Jbhd BEA JAocki 6 0 2 hs BEA Rockit 6 0 P2T 5 0 PRIMIERGY Tx150 Sp PRIMERGY T2150 x5 Xcon X321 P2T 4 0 P2T 4 0 P2T 2 0 Page 10 33 Fujitsu Technology Solutions 2009 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Benchmark environment The SPECjbb2005 measurements were performed on a PRIMERGY TX150 S6 with the following hardware and software configuration Hardware Meca S gegen d See oe 4 x 2 GB PC2 6400 DDR2 SDRAM Operating System Windows Server 2003 R2 Enterprise x64 Edition Xeon X3210 and X3360 BEA JRockit R 6 0 P27 4 0 build P27 4 0 10 90053 1 6 0 02 20071009 1827 windows x86 64 JVM Version Xeon X3370 BEA JRockit R 6 0 P27 5 0 build P27 5 0 5 97156 1 6 0 03 20080403 1524 windows x86 64 Fujitsu Technology Solutions 2009 Page 11 33 White Paper Performance Report PRIMERGY TX150 S6 SPECpower ssj2008 Benchmark description Version 5 1 November 2008 SPECpower ssj2008 is the first industry standar
17. 46 GB and 300 GB 15 krpm e 3 SATA hard disks with a capacity of 160 GB 250 GB 500 GB and 750 GB 7 2 krpm The hard disk cache has influence on disk I O performance Unfortunately this is frequently seen as a security problem in the event of a power failure and is therefore disabled On the other hand it was for a good reason integrated by the hard disk manufacturers to increase write performance Features such as Native Command Queuing NCQ only func tion at all when the disk cache is enabled For performance reasons it is advisable to enable the disk cache for the SATA hard disks in particular which in comparison with the SAS hard disks rotate slowly The by far larger cache for I O ac cesses and thus a potential security risk for data loss in the event of a power failure is in any case in the main memory and is administered by the operating system To prevent data losses it is advisable to equip the system with an uninter ruptible power supply UPS SATA RAID Onboard Controller ICH9R The following illustrations use 372 SATA hard disks to show how throughput depends on cache settings The through puts of a single hard disk Single Disk SD are compared with the throughputs of two hard disks in a RAID 0 and RAID 1 array Read throughput for sequential reading of 64 KB single Disk vs RAID 0 and RAID 1 with two Hard Disks blocks is not dependent on the cache settings In sequential Access RAID 1 roughly the same throughput values a
18. 5 disks Page 19 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 TX150 S6 with Xeon X3360 Fujitsu Technology Solutions 2009 80 x PRIMERGY BX300 2 x Pentium III 933 MHz 1 GB RAM 2 x Broadcom NetXtreme onboard Windows XP Professional SP1 PRIMERGY TX150 S6 1 x Xeon 3360 8 GB PC2 6400 DDR2 SDRAM 1 x Emulex LPe11002 fibre channel controller 2 x dual channel Intel PRO 1000GT 1 x Broadcom NetXtreme II BCM 5708 onboard Operating system Red Hat Enterprise Linux 5 1 2 6 18 53 el5 x86 64 HTTP software Accoria Rock Web Server v1 4 6 x86_ 64 Disk subsystem 2 x FibreCAT SX88 with 24 disks Page 20 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 StorageBench Benchmark description To estimate the capability of disk subsystems Fujitsu Technology Solutions defined a benchmark called StorageBench to compare the different storage systems connected to a system To do this StorageBench makes use of the lometer measuring tool developed by Intel combined with a defined set of load profiles that occur in real customer applications and a defined measuring scenario Measuring tool Since the end of 2001 lometer has been a project at http SourceForge net and is ported to various platforms and en hanced by a group of international developers lometer consists of a user interface for Windows systems and the so called dynamo which is avail
19. D 5 For sequential read access with 64 KB blocks and enabled disk cache the throughput is about 18 higher The throughput is about 6 7 fold higher for sequential write access with 64 KB blocks and enabled disk cache The disadvantage of RAID 10 compared with RAID 5 lies in the poorer capacity utilization The loss of capacity in a configuration with four hard disks is 50 with RAID 10 and only 25 with RAID 5 The throughput differences between RAID 5 and RAID 10 are also evident during random access However these differences are not as prominent as with sequential write The throughput difference between RAID 5 and RAID 10 depends on the block size For random access with 8 KB blocks the throughput in RAID 10 is 25 higher and with 64 KB blocks about 14 higher than in RAID 5 For random access the enabling of the disk cache brings about a throughput increase of between 27 and 39 with RAID 10 and about 24 with RAID 5 The slightly poorer throughputs measured in the RAID 5 array can be ex plained through the additional outlay required during the creation of the parity block Fujitsu Technology Solutions 2009 RAID 5 and RAID 10 with four Hard Disks random Access IBIS SKE random B4 KE random ETH reads BT reads Es IE 10 RAIDS RAID 10 OHO Cache aff cs ot El HO Cache on tm JS 0 589 669 Page 24 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 LSI MegaRA
20. HDs sequential Access maximum possible throughput of over 100 MB s The I O cached EES cache option has a negative impact on read throughput in RAID 1 TOP dj JJ AJ ee Ou LSI MegaRAID SAS 1078 with 512 MB Cache 2 5 15 krpm Hard Disks performance it is necessary to use anans 286 205 z27 227 269 269 z28 z27 10 0 na 100 n4 280 286 294 SE c write through WO cached Disk cache disabled g Write back PO cached Disk cache disabled shows that sequential write through Write back and achieves even higher values than with sequential read although an additional parity block has to be In contrast the write throughput is more dependent on the cache set tings In order to achieve optimum 64 KE sequential reads 64 KE sequential writes the Disk cache enabled option as a d d leit h the optimum cache setting The RA Banoo ne E NUT Hem st oe acne settings The importance of optimal cache a Write through WO direct Disk cache disabled e Write back UO direct Disk cache disabled Settings can be seen particularly b Write through MO direct Disk cache enabled Fo Write back WO direct Disk cache enabled clearly with RAID 5 The diagram d Write through PO cached Disk cache enabled h write back PO cached Disk cache enabled put increases considerably by a factor of 30 as a result of enabling the controller cache with the option calculated and written for write On the oth
21. ID SAS 1068 Below is a comparison of the performance of various hard disk types with the LSI MegaRAID SAS 1068 controller The controller does not have a controller cache Therefore measurements were only performed with and without a disk cache In the test setup two hard disks were connected to the controller and configured as a RAID 1 In the measurements all the hard disk types currently available to the PRIMERGY TX150 S6 were analyzed The throughputs of the individual hard disks types in RAID 1 are compared below with different access patterns The diagram shows that as the rotational speed increases the throughput for sequential reads and writes with a 64 KB block size rises RAID 1 with two Hard Disks sequential Access IES 70 23 79 40 114 72 If for sequential read with enabled disk cache a hard disk with a rotational speed of 15 krpm is used instead of one with a speed of 10 krpm the result for the 272 hard disk is an increase in throughput of about 27 and about 43 for the 372 hard disk If you compare the throughputs of the 272 and 372 hard disks both with a rotational speed of 10 krpm you can see that the throughput for the 372 hard disk is about 9 higher than for the 272 hard disk At a rotational speed of 15 krpm the difference in throughput between the Ze and 377 hard disk is even greater and amounts to 23 If you compare the 377 SAS hard disk with the 377 SATA hard disk you can then see that the through
22. Processor Celeron 440 U BH 2 2 2 Core 2 Duo E4500 2 2 2 2 2 2 4 4 4 4 4 e 0 60 80 8E All results were determined on the basis of the operating system Microsoft Windows Server 2008 Enterprise x64 Edition and the database SQL Server 2008 Enterprise x64 Edition OLTP 2 benchmark results depend to a great degree on the configuration options of a system with hard disks and their controllers Therefore the system was equipped with two dual channel Fibre Channel controllers that were connected to a total of 180 hard disks via two FibreCAT CX500 See the Benchmark environment section for further information on the system configuration The diagrams below show the OLTP 2 performance data for PRIMERGY TX150 S6 separated in two groups Celeron Pentium Dual Core and Core 2 Duo processors and a second group with Xeon processors There is a high increase of 115 at Celeron to Pentium Dual Core related to doubling the number of cores and the L2 cache size The scaling over all other processor types is about 3 to 25 and depends on processor and front side bus frequency increase larger number of cores and bigger L2 cache B 5 3 0 16 1 B 0 3 3 0 3 0 4 6 7 3 0 8 0 8 Fujitsu Technology Solutions 2009 Page 30 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 175 4 125 100 75 5 2 OLTP 2 PRIM ERGY TX150 S6 tps
23. SAS 1068 controller are not just better performance during random access to the SATA hard disks but also greater flexibility and scalability The LSI SAS 1068 controller supports more RAID levels and 372 and 272 SAS hard disks with 10 krpm and 15 krpm Thanks to the higher rotational speed 15 krpm of the SAS hard disks throughputs increases of between about 5596 and 193 can be achieved with all access modes in comparison with the SATA hard disks 7 2 krpm The user must decide for himself whether his needs are best covered by a less expensive solution with lower performance or a more expen sive higher performance solution Conclusion With the Modular RAID concept the PRIMERGY TX150 S6 offers a plethora of opportunities to meet the various re quirements of different application scenarios The entry level controller represented by the LSI MegaRAID SAS 1068 controller offers the basic RAID solutions RAID 0 RAID 1 and RAID 1E and supports these RAID levels with a very good performance The high end controller represented by the LSI MegaRAID SAS 1078 controller offers all today s current RAID solu tions for the PRIMERGY TX150 S6 which can be expanded with up to eight internal hard disks this can be RAID levels 0 1 5 6 10 50 and 60 This controller is supplied with a 256 MB or 512 MB controller cache and can as an optional extra be secured with a BBU Various options for setting the use of the cache enable controller performan
24. able for various platforms For some years now it has been possible to download these two components under Intel Open Source License from http www iometer org or http sourceforge net projects iometer lometer gives you the opportunity to reproduce the behavior of real applications as far as accesses to IO subsystems are concerned For this purpose you can among other things configure the block sizes to be used the type of access such as sequential read or write random read or write and also combinations of these As a result lometer provides a text file with comma separated values csv containing basic parameters such as throughput per second transactions per sec ond and average response time for the respective access pattern This method permits the efficiency of various subsys tems with certain access patterns to be compared lometer is in a position to access not only subsystems with a file sys tem but also so called raw devices With lometer it is possible to simulate and measure the access patterns of various applications but the file cache of the operating system remains disregarded and operation is in blocks on a single test file Load profile The manner in which applications access the mass storage system considerably influences the performance of a storage system Examples of various access patterns of a number of applications Application Access pattern Database data transfer random 67 read 33 write 8 KB SQL Ser
25. adings for reporting The adjoining diagram gives an overview of the different components of this framework SPEC SPECpower_ssj2008 and the SPEC logo are registered trademarks of the Standard Performance Evaluation Corporation SPEC Fujitsu Technology Solutions 2009 Page 12 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Benchmark results In June 2008 the PRIMERGY TX150 S6 was measured with an Intel Xeon X3360 processor and 4 GB of PC2 6400E DDR2 SDRAM memory The measurement was taken under Windows Server 2003 R2 Enterprise x64 Edition and a JRockit R 6 0 P27 5 0 JVM by Oracle was used With the Xeon X3360 processor the PRIMERGY TX150 S6 achieved a world record score which exceeded the previous front runner by 6 7 in energy efficiency Compared to the measurement on the IBM System x3200 M2 which was per formed with the same processor and achieved nearly the same throughput in ssj ops this advance of the SPECpower_ssj2008 result of the PRIMERGY TX150 S6 is explained only by the lower power consumption at all load levels Additional tests with different configurations have been performed to show their influence on the server efficiency Fujitsu does not submit all SPECpower_ssj2008 measurements for publication at SPEC So not all the results presented here appear on SPEC s web sites But because we archive the results and log data for all measurements we are able to prove the correct
26. ance Evaluation Corporation SPEC Fujitsu Technology Solutions 2009 Page 8 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Benchmark results In November 2007 the PRIMERGY TX150 S6 was measured with the Xeon X3210 processor and a memory of 8 GB PC2 6400 DDR2 SDRAM The measurement was taken under Windows Server 2003 R2 Enterprise x64 Edition SP1 As JVM two instances of JRockit R 6 0 P27 4 0 build P27 4 0 10 90053 1 6 0 02 20071009 1827 windows x86 64 by BEA were used In January 2008 the PRIMERGY TX150 S6 was measured in an otherwise unchanged configuration with the Xeon X3360 processor With the Xeon X3360 the PRIMERGY TX150 S6 achieved the best result of all mono processor servers with a Quad Core processor and exceeded the previous front runner in this category by 2396 With the measurement of the PRIMERGY TX150 S6 all the measured values between 2 and 4 warehouses were incorporated in the overall bench mark result With the measurement of the PowerEdge R200 this applies to all measured values between 4 and 8 ware houses SPECjbb2005 bops SE PRIMERGY TX150 56 vs Dell PowerEdge R200 bopz PRIMERGY Tx150 6 Xeon S60 2 vhs BEA JAockit 6 0 PaT A 0 om hd T Oooo 10000 e ajy Dell PowerEdge R200 Keon X323 1 JV M BEA JEockit 6 0 P2T 4 0 SOO Warehouses SPECjbb2005 bops PRIMERGY TX150 56 vs Dell PowerEdge R200 200000 bopz 000i T0220 Tanon
27. begins During the measuring phase all requests and responses are recorded in the final results In the ramp down phase which now follows the threads are stopped followed by an idle phase before the next test iteration begins with another ramp up phase Thus altogether three iterations are performed for each workload The number of generated threads is defined separately for each workload according to the performance of the SUT in the test configuration To determine the results the clients measure for each requested page the time between the sending of the request and the arrival of all the data of the requested page The response times for embedded image files are also included in the calculation The result takes all those pages into account that meet particular QoS Quality of Service criteria For this purpose the responses are assigned to the following categories according to response times Banking and Ecommerce and transfer rates Support within the workloads GOOD response time lt 2s Banking lt 3s Ecommerce transfer rate gt 99000 bytes s Support TOLERABLE response time lt 4s Banking lt 5s Ecommerce transfer rate gt 95000 bytes s Support FAILED response time gt 4s Banking gt 5s Ecommerce transfer rate 95000 bytes s Support In all three test iterations at least 95 of all responses must fall into category GOOD and 99 into category TOLERABLE for the workload result to be valid A regular
28. ce to be flexibly adapted to suit the RAID levels used Use of RAID 5 or RAID 6 enables the existing hard disk capacity to be utilized economically for a good performance However we recommend a RAID 10 for optimal performance and security When connecting SATA hard disks in a RAID 0 RAID 1 or RAID 10 array it is best to use the onboard SATA RAID con troller The throughputs are comparable with those of LSI MegaRAID SAS 1068 1078 controllers and the higher CPU load on the SATA controller is easily catered for by modern processors However with RAID 5 the missing controller cache of the onboard SATA controller becomes particularly evident during sequential write If importance is attached to optimal performance then an LSI MegaRAID SAS 1078 controller should be chosen for SATA hard disks The PRIMERGY TX150 S6 offers a choice between SATA and SAS and for SAS hard disks between 272 hard disks and 372 hard disks and also different rotational speeds of 10 krpm or 15 krpm Depending on the performance required a decision must be taken as to which hard disk type with which rotational speed is to be used Hard disks with 15 krpm offer an up to 39 better performance As a result of using 272 hard disks it is possible depending on the RAID level to achieve higher parallelism through the use of more hard disks in the RAID array For maximum performance it is advisable particularly with SATA hard disks or when using a controller without a contr
29. chmark is available as source code and is compiled before the actual benchmark Therefore the compiler version used and its optimization settings have an influence on the measurement result SPECcpu2006 contains two different methods of performance measurement The first method SPECint2006 and SPECfp2006 determines the time required to complete a single task The second method SPECint_rate2006 and SPECfp_rate2006 determines the throughput i e how many tasks can be completed in parallel Both methods are addi tionally subdivided into two measuring runs base and peak which differ in the way the compiler optimization is used The base values are always used when results are published the peak values are optional Compiler eer SPECint2006 integer peak aggressive Speed single threaded SPECint_base2006 integer base por MN SPECint rate2006 aggressive throughput multithreaded SPECint rate base2006 SPECfp2006 floating point peak aggressive speed single threaded SPECfp_base2006 floating point base conservative SPECfp_rate2006 floating point peak aggressive throughput multithreaded SPECfp_rate_base2006 floating point base conservative The results represent the geometric mean of normalized ratios determined for the individual benchmarks Compared with the arithmetic mean the geometric mean results in the event of differingly high single results in a weighting in favor of the lower single results
30. co THE POSSIBILITIES ARE INFINITE FUJITSU WHITE PAPER Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Pages 33 Abstract This document contains a summary of the benchmarks executed for the PRIMERGY TX150 S6 The PRIMERGY TX150 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 envi ronment Contents jq BR os ELM 2 PE 3 EE E 8 cu be d 6i D L eed 12 EE R 17 MORI ate che HT INE 21 9M f 30 Boc m M 33 deir eR RP T 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Technical Data The PRIMERGY TX150 S6 is a mono socket tower server It includes the Intel 3210 chipset one Celeron Pentium Dual Core Core 2 Duo Dual Core Xeon or Quad Core Xeon processor up to 8 GB PC2 6400 DDR2 SDRAM depend ing on the processor used a 800 MHz 1067 MHz or 1333 MHz f
31. cutes the operations on these objects The business operations are assumed by TPC C e New Order Entry e Payment e Order Status Inquiry e Delivery e Stock Level Supervision e 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 e bops business operations per second is the overall rate of all business operations performed per second e bops JVM is the ratio of the first metrics and the number of active JVM instances In comparisons of various SPECjbb2005 results it is necessary to state both metrics The following rules according to which a compliant benchmark run has to be performed are the basis for these 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 MaxWhm but not less than 8 warehouses MaxWhm is the number of warehouses with the highest opera tion rate per second the benchmark expects Per default the benchmark equates MaxWH with the number of CPUS visi ble by the operating system The metrics bops is the arithmetic average of all measured operation rates with between MaxWhm warehouses and 2 MaxWhm warehouses SPEC SPECjbb and the SPEC logo are registered trademarks of the Standard Perform
32. d SPEC benchmark that evaluates the power and performance charac teristics of server class computers With SPECpower ssj2008 SPEC has defined server power measurement standards in the same way they have done for performance The benchmark workload represents typical server side Java business applications The workload is scalable multi threaded portable across a wide range of operating environments and economical to run It exercises CPUs caches memory hierarchy and the scalability of symmetric multiprocessor systems SMPs as well as implementations of the Java Virtual Machine JVM Just In Time JIT compiler garbage collection threads and some aspects of the operating system SPECpower ssj2008 reports power consumption for servers at different performance levels from 100 percent to active idle in 10 percent segments over a set period of time The graduated workload recognizes the fact that processing loads and power consumption on servers vary substantially over the course of days or weeks To compute a power performance metric across all levels measured transaction throughputs for each segment are added together and then divided by the sum of the average power consumed for each segment The result is a figure of merit called overall ssj ops watt This ratio gives information about the energy efficiency of the measured server Because of its defined measurement standard it allows the customers to compare it to ot
33. decessor the PRIMERGY TX150 S5 both in their highest per formance configurations an increase is noted in the integer test suite of 127 with SPECint rate base2006 and 154 with SPECint rate2006 In the floating point test suite the increase is 92 with SPECfp rate base2006 and 102 with SPECfp rate2006 SPECcpu z0d6 PRIMERGY TX150 56 SPEC cpu2006 PRIMERGY TX150 56 vs PRIMERGY TX150 55 vs PRIMERGY TX150 55 GU 70 2U DU 40 zu 30 40 30 2n zu SPECint_rateZ00e up SPECFp_rate2006 10 SPECint_rate_base 200 SPECKp_rate_basc2006 U PRIMERGY Tx150 5 PRIMERGY T2150 56 PRIMERGY Tx150 5 PRIMERGY T2150 Cp Xeon SOTO Xeon wo PO Moon SOTO Xeon SAATA 2 corts 4 cores 2 corts d cores 2 67 GHz 3 GHz 2 67 GHz 3 GHz 1067 MHz FSB 1333 MH2 FSB 1067 hH2 FSB 1333 MHz FSB 4 ME Le cache 12 MIE Le cache 4 ME L2 cache 12 MIB Le cache B5 Matt 35 watt B5 Matt 35 watt Fujitsu Technology Solutions 2009 Page 6 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Benchmark environment All SPECcpu2006 measurements were performed on a PRIMERGY TX150 S6 with the following hardware and software configuration Hardware Celeron 440 Pentium Dual Core E2140 E2160 and E2200 Core 2 Duo E4500 E4600 and E7200 Xeon 3065 E3110 E3120 X3210 X3220 X3350 X3360 and X3370 Number of CPUs Celeron 440 I D on chip per chip Pentium Dual Core E2140 E2160 and E2200 I D on chip per chip Core 2 Duo E4500 a
34. e lowest power of all processors But the lower power consumption of this CPU can not compensate the enormous performance drawback which is caused by the fact that this CPU has only 2 cores per chip The Xeon X3360 processor shows the ideal balance between performance and power consumption and thus makes it the best choice for SPECpower ssj2008 The following diagram displays a comparison of additional configuration options It only shows the differences in power consumption for each load level but no performance changes The performance depends on the choice of the CPU as you could see in the diagram before All the additional configuration changes we made in the following comparison like adding more memory more hard disks other and additional Power Supply Units PSUs and turning off power manage ment features do not have any or only very limited influence on performance Fujitsu Technology Solutions 2009 Page 14 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 SPEC power sei A0 PRIMERGY TX150 S6 configuration comparison 710 AER x yf ge S 129 Co x pawer watts 100 em Set 70 rs 50 40 30 20 10 active idle load level Yu Xeon X337Q0 S258 DIMMS A Xeon X33T0 Cp DIMMs 1xHDD r Keon XX370 20668 DIMMs HDD no PVR mom ix Red PSU Xeon 32570 4066 DIMMs zm FNIT mgr As you can see in the diagram above not only the right choice of processors
35. ed as the unit of measurement for the performance of the system measured 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 Technology Solutions The partially enormous hardware and time expenditure for standardized benchmarks has been reduced to a reasonable degree in OLTP 2 so that a variety of configurations can be measured within an acceptable period of time Even if the two benchmarks OLTP 2 and TPC E simulate similar application scenarios using the same workload 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 price performance calculation Benchmark results The PRIMERGY TX150 S6 has been released with a large number of processor types The following table is an over view which processors have been measured with the OLTP 2 benchmark uom rm ur 2 MB per chip 800 MHz 35 watt 1 MB per chip 800 MHz 65 watt 1 MB per chip 800 MHz 65 watt 20 1 MB per chip 800 MHz 65 watt 2 20 2MB per chip 800 MHz 65 watt 62 38 40 2 MB per chip 800 MHz 65 watt 3 MB per chip 1067 MHz 65 watt 4 MB per chip 1333 MHz 65 watt
36. entium Dual Core and Core 2 Duo processors Celeron Pentium DC Pentium DC 440 E2140 E2160 Pentium DC Core 2 Duo Core 2 Duo Core 2 Ouo E2200 E4500 E4600 ET2o0 SPECint rate2O068 F amp PECint rate base2006 Fujitsu Technology Solutions 2009 Page 4 33 White Paper Performance Report PRIMERGY TX150 S6 SPEC cpu2z006 PRIMERGY TX150 56 with Xeon processors DU TO DU 20 SPECint_rate2Qoe SPECint_rate_base 200 Xeon Xeon Xcon Xcon Xeon Xcon Xeon Xcon 2065 E110 E3120 A 2210 xs22n xs350 X33b5 EHET Version 5 1 November 2008 The measured SPECint_rate_2006 results are 12 17 above the SPECint_rate_base2006 results cores Processor e E o 6 BE 2 2 8 4 1 MB per chip 800 MHz 65 watt 0 0 0 0 0 3 3 d 4 MB per chip 1333 MHz 65 watt d 6 MB per chip 1333 MHz 65 watt 8 5 3 1 1 H 4 6 83 6 MB per 2 cores 1333 MHz 95 watt H 4 MB per 2 cores 1067 MHz 95 watt 7 3 0 D BRB RB BRIM MIMI MIM MLM NMI NM E SPECcpu20606 PRIMERGY TX150 56 with Celeron Pentium Dual Core and Core 2 Duo processors 3 Eu eu SPECKp_ratef006 SPECFp_rate_base2006 Celeron Pentium DC Pentium DC Pentium DC Core 2 Ouo Core 2 Duo Core 2 Oyo 440 E2140 E2160 E2200 E4500 E4600 ET 00 Fujitsu Technology Solu
37. er hand the cache settings have less impact on throughput with sequential read It is interesting to see how counterproductive the effect of enabling the I O cache is on read throughput particularly for reads Fujitsu Technology Solutions 2009 Page 26 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 The following graphic shows that during random access in RAID 1 the combination of the Write back and Disk cache enabled options has a decisive influence on throughput The improvement in throughput is 48 or 56 depending on RAID 1 2 HD and RAID 5 4 HDs random Access MBs Cache Settings a Write thraugh b Wwrite thraugh G Wwrite thraugh d Wwrite thraugh WC direct IC direct SEB random 57 reads a bj ec d e f g h a bio d f g h ORADI 23 27 26 27 28 35 26 27 t6 1614 ta t82 231 165 1 3 BRAIDS 26 30 27 30 42 42 41 41 153 164 15 3 16 2 19 6 235 224 21 3 Disk cache disabled Disk cache enabled HO cached Disk cache disabled WO cached Disk cache enabled whether access is with 64 KB or 8 KB block size For random access in RAID 5 it is advisable to enable the controller cache by setting the write mode option to Write back In order to achieve optimal throughput it is also necessary to enable the disk cache As a result of these optimal cache settings improvements in throughput of 54 or 60 can be achieved dependi
38. h the SATA hard disk by ena bling the disk cache The increase in throughput gained with SAS hard disks through enabling the disk cache is not so pronounced as with the SATA hard disks but it is still significant For the 277 hard disks with 10 krom throughput in creases by 44 and by about 62 for the 272 hard disks with 15 krpm For the 372 hard disks with 10 krpm throughput increases by 34 and by about 66 for the 372 hard disks with 15 krpm Fujitsu Technology Solutions 2009 Page 25 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 The following diagram shows that for random access with 67 reads the disk cache also plays an important role in im proving throughput RAID 1 with two Hard Disks random Access MES EE ze ze 20 ean woo wm mu GHOCacheon 1M 285 248 ata 322 a57 19 23 m42 am An increase in throughput of up to 30 was achieved for the SATA hard disk with 8 KB blocks For the 372 15 krom SAS hard disks the increase in throughput with 8 KB blocks is somewhat smaller namely 2396 If you compare the throughput of the 372 SAS hard disk with the 372 SATA hard disk you can then see that the throughput of the SAS hard disk with 10 krpm is about 2 2 times higher for random access with 8 KB blocks and enabled disk cache than in the SATA hard disk with 7 2 krpm If you compare the throughput of the 372 SAS hard disk with 15 krpm with the 377 SATA hard
39. her configurations and servers measured with SPECpower ssj2008 The adjoining diagram shows a typical graph of a SPECpower ssj2008 result target load SUT system Under Test CCS Control amp Collection System Any OS Linux Solaris Windows Control Collect e ccs properties PTDaemon PTDaemon power I temp Power Analyzer ssj 2008 JVM instance s ssj properties ssj2008 Director Temperature Sensor AC Power achive Performance to Power Ratio FOO Hal 1000 1250 1 124 owverallzsj opsiwatt 1 744 LEI 1434 1 354 p D 1500 1750 100 30 Bu TUM B DU 40 30 eum 10 idle 4 DU Bn T Average Power w The benchmark runs on a wide variety of operat ing systems and hardware architectures and does not require extensive client or storage infrastructure The minimum equipment for SPEC compliant testing is two networked com puters plus a power analyzer and a temperature sensor One computer is the System Under Test SUT running any of the supported operating systems along with the JVM installed The JVM provides the environment required to run the SPECpower ssj2008 workload which is imple mented in Java The other computer is a Collect and Control System CCS which controls the operation of the benchmark and captures the power performance and temperature re
40. ironment All SPECpower_ssj2008 measurements presented here were performed on a PRIMERGY TX150 S6 with the following hardware and software configuration using the ZES Zimmer LMG95 power analyzer Hardware Model PRIMERGY TX150 S6 Processor TDP Xeon E3120 65 W X3220 95 W X3360 95 W X3370 95 W 1 chip 2 cores per chip and 4 cores per chip 32 KB instruction 32 KB data on chip per core 8 MB I D on chip per chip and 12 MB I D on chip per chip 6 MB shared 2 cores 2x2 GB PC2 6400E DDR2 SDRAM 4 x 2 GB PC2 6400E DDR2 SDRAM Network Interface 1 x 1 GBit LAN Broadcom onboard 1 x Integrated SATA Controller Disk Subsystem 1 x 3 5 SATA disk 160 GB 7 2 krpm JBOD 4 x 3 5 SATA disk 160 GB 7 2 krpm JBOD 1 x 350 W DPS 350UB A Power Supply Unit 1 x 450 W DPS 450FB G 2 x 450 W DPS 450FB G Memory Operating System Windows Server 2003 R2 Enterprise x64 Edition Oracle JRockit R 6 0 P27 5 0 build P27 5 0 5 o CR371811 CR374296 100684 1 6 0 03 20080702 1651 windows x86 64 JUNLebtions Xms1650m Xmx1650m Xns1500m XXaggressive Xlargepages Xgc genpar XXcallprofiling H XXgcthreads 2 XXtlasize min 4k preferred 1024k XXthroughputcompaction JVM Version Fujitsu Technology Solutions 2009 Page 16 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Ht SPECweb2005 spec Benchmark description SPECweb2005 is the next generation web server benchmark developed by the O
41. nd E4600 I D on chip per chip Core 2 Duo E7200 I D on chip per chip RECON SACHE Xeon 3065 I D on chip per chip Xeon E3110 and E3120 I D on chip per chip Xeon X3210 and X3220 I D on chip per chip Xeon X3350 X3360 and X3370 I D on chip per chip 8 GB PC2 6400 DDR2 SDRAM B B B B B B B B Xeon 3065 SUSE Linux Enterprise Server 10 64 bit Pentium Dual Core E2200 Core 2 Duo E7200 Xeon E3120 and Xeon X3370 SUSE Linux Enterprise Server 10 SP2 64 bit others SUSE Linux Enterprise Server 10 SP1 64 bit Compiler Xeon E3120 Intel C Fortran Compiler 11 0 d others Intel C Fortran Compiler 10 1 Operating System Fujitsu Technology Solutions 2009 Page 7 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 SPECjbb2005 spec Benchmark description SPECjbb2005 is a Java business benchmark that focuses on the performance of Java server platforms It is essentially a modernized version of SPECjbb2000 with the main differences being e The transactions have become more complex in order to cover a greater functional scope e The working set of the benchmark has been enlarged to the extent that the total system load has increased e 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
42. ng on the block size used E4 KE random 577 reads e Write back WO direct Olek cache disabled Fo Write back PO direct Olek cache enabled g Write back WO cached Disk cache disabled bh Write back PO cached Disk cache enabled LSI MegaRAID SAS 1078 with 512 MB Cache 2 5 15 krpm Hard Disks More detailed information about this topic is available in the paper Performance Report Modular RAID for PRIMERGY Controller comparison The following comparison depicts the throughputs of the various controllers The measurements were made with the same hard disk types in the same RAID array The diagram shows the throughputs achieved with disabled caches Off and with optimal cache settings Optimal You can see that the cache set tings for sequential read access do not have any or only a very minor influence on throughput regardless of which controller is used The throughput values achieved are equivalent to the maximum possible values For sequential write access it is pos sible to achieve an increase in throughput by optimizing the cache settings Throughput in creases by about 66 with the LSI MegaRAID SAS 1068 con troller The best throughput val ues were achieved with the LSI MegaRAID SAS 1078 controller with a 512 MB controller cache The difference in performance to Controller LSI S z 1068 and LSI SAS 1078 with 256 512 MB Cache RAID 1 with two 3 5 15 krpm SAS Hard Disks IE 160 4 um
43. ol ler cache to enable the hard disk cache Depending on the disk type used and access pattern the increase in perform ance is 13 fold When the hard disk cache is enabled we recommend the use of a UPS Fujitsu Technology Solutions 2009 Page 28 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Benchmark environment All the measurements presented here were performed with the hardware and software components listed below Component pm Version 5 2 3790 Service Pack 1 Build 3790 Measuring tool lometer 27 07 2006 Measurement data Measurement file of 8 GB Product Intel ICH9R Onboard SATA RAID Driver Name megasr sys Driver Version 09 21 0914 2007 Product LSI RAID 0 1 SAS 1068 Controller LSI MegaRAID SAS 1068 Driver Name Isi_sas sys Driver Version 1 25 05 00 Product LSI RAID 5 6 SAS 1078 Driver name msas2k3 sys driver version 2 17 0 32 Controller LSI MegaRAID SAS 1078 Firmware package 6 0 1 0074 firmware version 1 11 32 0307 BIOS version NT10 Controller cache 256 MB or 512 MB Fujitsu Technology Solutions 2009 Page 29 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 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 determin
44. on processor comparison diagram shows measurements with four different Xeon CPUs measured with the PRIMERGY TX150 S6 All other configuration details remained unchanged during the measurements SPEC power s5j2008 Intel Xeon processor comparison 725000 125 ral tt 115 175000 5 105 150000 o e e 85 i 100000 rel e TEM 75 0 i 45 TOO a By TU ay Eat Ki As 20 ds 10 active ide bad asi 0 EE r el Xeon E3120 sel om E tel Xeon X3220 sel ops NENNEN r el Xeon X2360 sel opc Wm r el Xeon X3370 ssj ops throughput ssj_ops power watts i S inel Xeon X336 wai intel Xeon X3370 wai c inel Xeon E3120 wai infiel Xeon Was Obviously the throughput changes when using different CPUs with different frequencies or different number of cores per chip That is exactly what the bars show in the diagram above The measurement with the most powerful Xeon X3370 quad core 3 0 GHz CPU delivers the highest throughput and the measurement with the Xeon E3120 dual core 3 16 GHz CPU which has nearly the same frequency but half the number of cores per chip delivers the lowest throughput in ssj ops left y axis It is also visible that the ratio of the performance gain and the performance loss respectively with the different CPUs is nearly the same at each target load level x axis But looking at the average power consumption curves right y axis the behavior varies for the different load levels
45. overall result requires valid partial results in all three workloads with the same system configuration The individual results are named after the workloads and indicate the maximum number of user sessions that can be handled by the system under test with the QoS criteria being met They thus allow a system to be assessed under differ ent realistic conditions To calculate the overall result each partial result is related to a reference value then the geomet ric mean of these three values is calculated multiplied by 100 The overall result SPECweb2005 thus indicates the relative performance of the measured system in relation to the reference system SPEC SPECweb and the SPEC logo are registered trademarks of the Standard Performance Evaluation Corporation SPEC Fujitsu Technology Solutions 2009 Page 17 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Benchmark results In April 2008 the PRIMERGY TX150 S6 was measured with one Xeon X3220 processor and 8 GB PC2 6400 DDR2 SDRAM Two quad port Intel PRO 1000GT and one Broadcom NetXtreme II BCM5708 onboard were used for the network A FibreCAT CX500 with 45 hard disks which was connected via an Emulex LP10000DC fibre channel control ler was used as disk subsystem A RAID 5 was built across the 45 hard disks The logging was done on a hard disk of type Seagate ST380013AS The operating system was resident on a Seagate ST3160815AS hard disk Both
46. pen Systems Group OSG of the Standard Performance Evaluation Corporation SPEC It is the successor of SPECweb99 and SPECweb99 SSL and it measures the performance of a HTTP server under a standardized load of static and dynamic requests The new version includes many sophisticated and state of the art enhancements to meet the modern demands of Web users of today and tomorrow Contrary to its predecessor version SPECweb2005 is split into three different workloads which are based on real world web server applications SPECweb2005 Banking Emulates typical online banking requests such as login logoff account status bank transfers displaying and changing user profiles etc Login includes the setting up an SSL connection that will be used for all following activities SPECweb2005 Ecommerce Simulates an online transaction in the computer business Users can look through the pages view goods put them in their shopping carts and purchase the products Activities in the initial phases of the connection use non encrypted connections As soon as an order is to be sent off the connections are SSL en crypted SPECweb2005 Support Emulates requests coming in on a support web site Users can search through the page view lists of available products and download the related files Requests are always non encrypted The requests of all three workloads refer to dynamically generated contents and static files of various sizes Intervals be
47. put of the SAS hard disk with 10 krpm is about 12 higher than the SATA hard disk with 7 2 krpm for sequential read and with an en abled disk cache If you compare the 377 SAS hard disk with 15 krpm with the SATA hard disk you see that the throughput of the 372 SAS hard disk with 15 krpm is even 609 higher than with the SATA hard disk If for sequential write with enabled disk cache a hard disk with a rotational speed of 15 krpm is used instead of one with a speed of 10 krpm the result for the 272 hard disk is an increase in throughput of about 26 and about 39 for the 372 hard disk If you compare the throughputs of the 277 and 372 hard disks both with a rotational speed of 10 krpm you can see that the throughput for the 372 hard disk is about 4 higher than for the 272 hard disk At a rotational speed of 15 krpm the difference in throughput between the Ze and 377 hard disk is even greater and amounts to 15 If you compare the 377 SAS hard disk with the 377 SATA hard disk you can then see that the throughput of the SAS hard disk with 10 krpm is about 1896 higher than the SATA hard disk with 7 2 krpm for sequential write and with an en abled disk cache If you compare the 372 SAS hard disk with 15 krpm with the same SATA hard disk you see that the throughput of the 372 SAS hard disk with 15 krpm is even 63 higher than with the SATA hard disk A special increase in throughput for sequential write up to 10 4 fold can be achieved wit
48. put were found in m sequential writing and deactivated disk cache In RAID 1 witht AG fk SATA Hard Disk diii Lal uiuis this access pattern the onboard SATA controller is MBs 44 faster than the LSI MegaRAID SAS 1068 con Ba sa troller however the absolute throughput values in 100 DU both cases are less than 10 MB s while the activa A tion of the disk cache on both controllers leads to an increase in performance by a factor of 7 or 10 to 70 MB s E SECH o a During random access with enabled disk cache on Onboard SATA E TER SAS 1068 the other hand the LSI MegaRAID SAS 1068 con a b troller offers about 18 more performance than the EE ACER BENE Ear og onboard SATA controller BHO Cache on 77 46 7026 0 97 T24 7797 E ETE The onboard SATA controller is implemented di rectly on the motherboard of the server in the Intel Access Pattern ICH9R chipset The RAID stack is handled by the Scenes SE gat e Gig server CPU The increased load on the CPU de c Database random SKB BT read 3354 write pends on the access pattern and block aei The d File Server random 64KB EX read 23 write CPU load increases by up to 5 percentage points in this example although the CPU load in many small data blocks is generally higher Although the on board SATA controller features good performance figures the benefits of the LSI SAS 1068 controller should not go unmentioned The advantages of the LSI
49. r LSI MegaRAID SAS 1068 The controller is supplied as a PCI Express card The maximum number of SATA and SAS hard disks that can be connected to the controller is eight Support is provided for RAID levels 0 1 and 1E The controller does not have a cache 3 RAID Controller LSI MegaRAID SAS 1078 The controller is supplied as a PCI Express card and offers the user a complete RAID solution Both SATA and SAS hard disks can be connected Support is provided for RAID levels 0 1 5 6 10 50 and 60 Two different versions of this controller are on offer with either a 256 MB or 512 MB cache The controller cache can be pro tected against power failure by an optional battery backup unit BBU The controller supports up to 240 hard disks Various SATA and SAS hard disks can be connected to these controllers Depending on the performance required it is possible to select the appropriate disk subsystem Depending on the model version the PRIMERGY TX150 S6 offers four 3v SAS SATA or eight 272 SAS hot plug bays for hard disks Optionally an extension box is available for the 372 variant with two additional 372 hot plug bays The following hard disks can be chosen for the PRIMERGY TX150 S6 e 2 SAS hard disks with a capacity of 36 GB 73 GB and 146 GB 10 krpm e 2 SAS hard disks with a capacity of 36 GB and 73 GB 15 krpm e 3 SAS hard disks with a capacity of 73 GB 146 GB and 300 GB 10 krpm e 3 SAS hard disks with a capacity of 73 GB 1
50. re ME achieved as in the single disk configuration e e although RAID 1 offers the benefit of data re dundancy RAID O has a better utilization of capacity and two hard disks RAID O0 almost dou ble the read throughput In contrast write throughput with sequential access with 64 KB blocks largely depends on the cache settings Enabling the disk cache im proves the write throughput by a factor of about 11 in a single disk configuration by a factor of 13 in RAID O and by a factor of 7 in RAID 1 The much higher write throughput is explained by the optimized write accesses to the hard disk and 22 B se xm RAIO 0 SD RAID O RAD mHDCacheon 7242 win 7746 __ 7737 12943 7028 Fujitsu Technology Solutions 2009 Page 23 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 the shorter latency times Here again the RAID O array achieves almost twice the write throughput through parallel ac cesses aS compared with the other two configurations Enabling the disk cache leads to an increase in throughput during random access However this increase is not as noticeable as with sequential Single Disk vs RAID 0 and RAID d with two Hard Disks random Access writing With a random access with 64 KB blocks and a single disk configuration the increase in throughput is about 36 in RAID O about 22 and in RAID 1 about 19 In the case of random access with 8 KB blocks
51. ront side bus a Broadcom BCM5755 1 GBit LAN con troller six PCI slots 2 x PCI Express x8 1 x PCI Express x4 3 x PCI 32 bit 33 MHz and space for four 3 5 SAS or SATA hard disks or up to 10 2 5 SAS hard disks The SAS version of the PRIMERGY TX150 S5 has an 8 port SAS controller with RAID 0 1 and RAID 1E functionality or alternatively an 8 port SAS controller with RAID O 1 10 5 50 6 and RAID 60 functionality The SATA version has a 6 port SATA controller with RAID 0 1 10 and optionally RAID 5 functionality As well as its predecessors the PRIMERGY TX150 S6 can be converted quickly and easily into a rack system with 5 height units for integration in 19 inch racks E E i E Hi _ See http docs ts fujitsu com dl aspx id af1d8493 dfbb 4 78e 89c9 92e91 7929234 for detailed technical information Fujitsu Technology Solutions 2009 Page 2 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 SPECcpu2006 spec Benchmark description SPECcpu2006 is a benchmark to measure system efficiency during integer and floating point operations It consists of an integer test suite containing 12 applications and a floating point test suite containing 17 applications which are extremely computing intensive and concentrate on the CPU and memory Other components such as disk I O and network are not measured by this benchmark SPECcpu2006 is not bound to a specific operating system The ben
52. s of the Standard Performance Evaluation Corporation SPEC Fujitsu Technology Solutions 2009 Page 3 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 e Xeon E3110 and E3120 Wolfdale 2 cores per chip 6 MB L2 cache per chip e Xeon X3210 and X3220 Kentsfield 4 cores per chip 4 MB L2 cache per 2 cores e Xeon X3350 X3360 and X3370 Yorkfield 4 cores per chip 6 MB L2 cache per 2 cores The following two tables show results in which all benchmark programs were compiled with the Intel C Fortran com piler 10 1 See the Benchmark environment section fort he operating system versions used Bold result numbers are published at http www spec org Processor Pentium Dual Core E2160 AJAJ ISIkvba ib ibh ibh ib vb ibh ib lbh GHz i2cwhe FSB TDP SPECint rate base2006 SPECint rate2006 160 1 MB perchip 800MHz G5wat 183 Ia 1 80 1 MB per chip 800 MHz 65 watt 20 1 22 5 220 2MB perchip mus oswa 250 280 2 40 2MBperchip 800 MHz 65wat wu 295 233 AMBperchip 1333 MHz oswa 293 se B eMBperchip mus eswat 8 mn 213 4 MB per 2 cores 1067MHz oSwat 479 543 240 4MB por 2 cores 1067MHz 5wat s24 s2 2 67 6 MB per 2 cores 1333 MHz Ise 612 nmo 2 83 MB per 2 cores 1333 MHz o5wat eao ms B 6 MB per 2 cores 1333 MHz o5watt 662 me SPECcpu z006 PRIMERGY TX150 56 with Celeron P
53. the same time consume less power than a comparable performance equiva lent 4 DIMMs configuration We used the SATA base unit which has an integrated SATA controller in the chipset along with a 160 GB 3 5 SATA hard disk Due to the absence of a dedicated onboard SAS controller which is included in the SAS base unit and the lower rotational speed of 7 200 rpm of the SATA hard disk this configuration was the best choice for this benchmark because it consumes minimal power without impacting the performance The most important factor in the hardware configuration is the right choice of the processor Processors are the part of a server that consumes the most power beside the memory subsystem For the PRIMERGY TX150 S6 the quad core Xeon X3360 processor with a Thermal Design Power TDP of 95 watts showed the most efficient result The influence of the different CPUs and other configuration options on the SPECpower_ssj2008 result is presented in subsequent diagrams Competitive benchmark results stated above reflect results published as of Jun 4 2008 The comparison presented above is based on the most efficient servers currently shipping by IBM and Fujitsu Siemens Computers now operating under the name of Fujitsu For the latest SPECpower ssj2008 benchmark results visit http www spec org power ssj2008 results Fujitsu Technology Solutions 2009 Page 13 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 The Xe
54. tions 2009 2 MB per chip 800 MHz 65watt 234 3MBperchip 1067 MHz 65watt 267 392 MB perchip 1333 MHz Lee 310 ma a MB per2 cores 1383 MHz 95wot aas 465 BMBper2cwes 1383 MHz 95won 478 ee FSB TP s SPECf p rate base2006 SPECfp rate2006 2 MB per chip 800 MHz 35 watt 1MBperchip 800MHz 65watt 180 1 MB per chip 800 MHz 65 watt 2 2 MB per chip 800 MHz 65 watt 2 A co o Q3 NO N mh A ed Ew ed Ere VININ N gt Page 5 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 SPEC cpu2006 PRIMERGY TX150 56 with Xeon processors zu 40 AU eu SPECFp rarez e SPECFp_rate_base2006 Xcon Xcon Xeon Xeon Xeon Xcon Xeon Xcon S065 E3110 E3120 A 2210 sosi xs350 X335 0 waat The measured SPECfp rate 2006 results are 4 9 above the SPECfp rate base2006 results The following four tables show results in which all benchmark programs were compiled with the Intel C Fortran com piler 10 1 and run under SUSE Linux Enterprise Server 10 64 bit Significantly higher results were achieved with the Intel C Fortran compiler 11 0 than would have been expected with version 10 1 In principle this should also apply for other processors of the PRIMERGY TX150 S6 Xeon E3120 3 17 6 MB perchip 1333 MHz 65 watt L2 cache me TOP SPECfp_rate_base2006 SPECfp rate2006 When comparing the PRIMERGY TX150 S6 and its pre
55. tween requests think times vary The distribution of the requests and the think times are controlled by tables and functions Average values for these parameters are laid down in configuration files and are monitored by the sequencing unit SPECweb2005 is not tied to a particular operating system or to a particular web server The benchmark environment consists of several components Each client system runs a load generator program setting up connections to the web server sending page requests and receiving web pages in response to the requests A prime client initializes the other systems monitors the test procedure collects the results and evaluates them The web server also referred to as Sys tem Under Test SUT comprises the hardware and software used to handle the requests A new feature is the back end simulator BeSim that emulates the database and application components of the entire application The web server communicates with the BeSim via HTTP requests to obtain any additional information required The sequencer and the client programs are written in Java and are divided into individual threads each of which emulates a virtual user session All three workloads pass various phases during the test In the ramp up phase the load generating threads are started one after another This is followed by a warm up phase initializing the measurement Any previously recorded results and errors are deleted before the actual measuring interval
56. ust 5 2008 Compared with the PRIMERGY TX150 S5 the PRIMERGY TX150 S6 improved the throughput performance by 133 SPECweb2005 PRIMERGY TX150 56 vs PRIMERGY TX150 55 FRIFIERGTY Tx150 36 Xeon X335 a GB PC2 6400 ODR2 S0RAM PRIMERGT T2150 36 Xeon X322 a GB PC2 6400 DDRB2 SDRAM PRIMERGT T2150 35 eon 305r a GB PC2 4200 ODR2 S0RAM S000 LuLu LEDa DOu0 SZUDUU Competitive benchmark results stated above reflect results published as of August 5 2008 The comparison presented above is based on the best performing servers with one Quad Core processor currently shipping by Dell and Fujitsu Siemens Computers now operating under the name of Fujitsu For the latest SPECweb2005 benchmark results visit http www spec org web2005 results Fujitsu Technology Solutions 2009 Page 18 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Benchmark environment TX150 S6 with Xeon X3220 Fujitsu Technology Solutions 2009 60 x PRIMERGY BX300 2 x Pentium Ill 933 MHz 1 GB RAM 2 x Broadcom NetXtreme onboard Windows XP Professional SP1 PRIMERGY TX150 S6 1 x Xeon 3220 8 GB PC2 6400 DDR2 SDRAM 1 x Emulex LP10000DC fibre channel controller 2 x dual channel Intel PRO 1000GT 1 x Broadcom NetXtreme II BCM 5708 onboard Operating system Red Hat Enterprise Linux 5 1 2 6 18 53 el5 x86 64 HTTP software Accoria Rock Web Server v1 4 6 x86_ 64 Disk subsystem 1 x FibreCAT CX500 with 4
57. ver Database log file sequential 100 write 64 KB blocks Restore sequential 100 write 64 KB blocks Video streaming sequential 100 read blocks 2 64 KB File server random 67 read 33 write 64 KB blocks Web server random 100 read 64 KB blocks Operating system random 40 read 60 write blocks 2 4 KB File copy random 50 read 50 write 64 KB blocks From this four distinctive profiles were derived Load profile Access pattern Block Load All four profiles were generated with lometer Fujitsu Technology Solutions 2009 Page 21 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 Measurement scenario In order to obtain comparable measurement results it is important to perform all the measurements in identical repro ducible environments This is why StorageBench is based in addition to the load profile described above on the follow ing regulations Since real life customer configurations work only in exceptional situations with raw devices performance measurements of internal disks are always conducted on disks containing file systems NTFS is used for Windows and ext3 for Linux even if higher performance could possibly be achieved with other file systems or raw devices Hard disks are among the most error prone components of a computer system This is why RAID controllers are used in server systems in order to prevent data loss through hard disk failure Here several hard
58. y men tioned configurations It gives an overview of the benchmark result in overall ssj_ops watt left y axis each configuration achieved and how much power in watts right y axis was consumed while running the benchmark Energy efficiency PRIMERGY TX150 56 configuration compar son 1000 2000 1750 1000 1500 E ND 1250 a 5 600 539 1000 z au Dou 5 di B 400 E 50 200 f f 250 d THD AxHDD AHD 40D AxHDD ma PAR mami na PA mami no PWR mgm d Sa PSU ixRed PSU x Red PSU Fujitsu Technology Solutions 2009 Page 15 33 White Paper Performance Report PRIMERGY TX150 S6 Version 5 1 November 2008 As you can see the most energy efficient configuration does not consume the lowest power It is still about 108 more efficient compared to the least energy efficient configuration The configuration with the Xeon E3120 processor has the lowest power consumption Although the best configuration with the Xeon X3360 processor does not deliver the highest throughput in ssj ops and does not have the lowest power consumption it achieves the highest benchmark score of 1 124 overall ssj ops watt The other results are important too as they show the dependencies between the configurations and the efficiency This information should give some hints about the power consumption and efficiency that can be expected from real world configurations used in customer installations Benchmark env
Download Pdf Manuals
Related Search