Dell PowerEdge M915 Technical White Paper
Contents
1. Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Contents Executive SUMMANY vieres ses oves n e V veros e ee E eo VPVre rss V Vere oPke sve PU Pre VU PP eS EE E EVE ET MEFORUCHON EET Key FindIngs iod MI Seana sae aa ad a ae ae eM e ed ee PEFFOFMANCEY Watts cevscvescavsivessevs senses send deca rona cows ewes ces coud sensed GAEE aces deus N EEEE AEN Performan CO ia PPP Hoa Q MONO dOlO SW e A oA ERI REESE NU MIT m Results 21e rene ode IRI re Duodecim eed b ea os PL eei MID PETI Rea des Price as TeSte e Price to Power Efficiency Ratio ssissscssccsseussonssecetanssavssasacatecghiseds s doge eo Pa dete cons Se RP Eo le ee Pide sre Sss Performance per Dollar 2 cero creo pee reos e ee o geo ree gee eso gro reed Use eee neo ee lse eee dee coelis Industry Leading 4 Socket Server Power Efficiency c eeeeceeee esses esee eene eene hene eren hehehe hne nnn Future Platform Updates 1 oiccesecc esee e see eae sese s sens aden ae ose n sen een deno co p sens ern ses coa aen cra enu SUM MALY iiss eeevesse eee vescesecvescesveveseeveccvcvescevesvecversccesvezesenseeevevesvesceevecevsreceseeveevececececeeevvvseeVue Appendix A Test Methodology o oon e eeu eva dada ive ca a esya suae dv ENAERE AANTEKEN EN2. The 8 node M915 achieved 14 982 232 ssj_ops 100 target load and a SPECpower_ssj2008 result of 2 425 overall ssj ops Watt compared to the 4 node BL680c G7 with 7 996 829 ssj_ops 100 target load and a SPECpower ssj2008 result of 1 260 overall ssj_ops Watt 14 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Price as Tested Data in the previous section shows that the PowerEdge blade solution with M915 blades is more power efficient and higher performing that the ProLiant blade solution with BL680c G7 blades Despite having twice as many blade servers the Dell solution costs 21 less as tested Figure 6 Figure 6 Price as Tested Blade Solution Price as Tested U S Dollars lower is better 160 000 00 f lj 150 000 00 137 980 00 106 704 00 5 M915 and M1000e prices in U S dollars from Dell com June 14 2011 BL680c G7 and C7000 prices in U S dollars from HP authorized reseller June 14 2011 15 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Price to Power Efficiency Ratio Table 2 and Table 3Table 3 detail the price for both solutions including the blades blade enclosures chassis controller modules I O modules and power supplies necessary for the testing Table 2 Dell Blade Solution Cost Breakdown Dell Blade Solution Qty Each Total PowerEdge M915 blades 8 13 338 00 106 704 00 M1000e Modular Blade Enc
3. target utilization The Dell blade solution has twice as many servers as the HP solution but as Figure 4 shows the Dell solution consumes just 14 more power at full utilization Figure 4 Blade Solution Power Consumption at 100 Target Load Blade Solution Power Consumption at 10096 Target Load Despite providing twice as many servers and three times as many physical processor cores the Dell blade solution consumes just 1496 more power at 10096 utilization than the HP blade solution The 8 node M915 consumed just 4 573 W at 100 target utilization compared to 4 020 W for the 4 node BL680c G7 The 8 node M915 achieved 14 982 232 ssj_ops 100 target load and a SPECpower_ssj2008 result of 2 425 overall ssj_ops Watt compared to the 4 node BL680c G7 with 7 996 829 ssj_ops 100 target load and a SPECpower ssj2008 result of 1 260 overall ssj_ops Watt 13 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions SPECpower_ssj2008 also includes a measurement of power while the servers are at 0 target utilization Active Idle Despite having twice as many blade servers the Dell solution consumes 799 W less power at idle than the HP blade solution as seen in Figure 5 Figure 5 Blade Solution Power Consumption at Idle Blade Solution Power Consumption at Idle Lower is better 7 The 8 node M915 consumed only 1 492W at the Active Idle Point compared to 2 291W for the 4 node BL680c G7
4. DIMMs and CPUs the blades could accommodate The configuration used is summarized in Table 1 Table 1 Detailed Blade Solution Configurations Blade Component Sockets Form Factor Blades Per Solution Processors Per Blade Processors Per Solution Physical Processor Cores Per Solution Logical Processor Cores Per Solution Memory Slots Per Blade DIMMs Per Blade DIMMs Per Solution Hard Drives Per Blade Hard Drives Per Solution Storage Controller Chassis Components Management 1 0 Power Supply Quantity Rating Dell PowerEdge M915 4S Full Height Single Width 8 4 x AMD Opteron 6176 2 3 GHZ 12 Cores each 32 384 384 32 32 x 4GB Dual Ranked PC3L 10600R LV RDIMMs running at 1066MHz 256 2 x 146GB 15k 6Gb RAID 1 16 Dell PERC H200 No Cache Dell PowerEdge M1000e Modular Blade Enclosure 1 x Dell CMC Module 1 x Dell Ethernet Pass Through Module 6 x 2700W Platinum Rated Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions HP Proliant BL680c G7 4S Full Height Double Width 4 4 x Intel Xeon X7560 2 26 GHz 8 Cores 16 Threads each 16 128 256 64 64 x 2GB Dual Ranked PC3 10600R RDIMMs running at 1066MHz 256 2 x 146GB 15k 6Gb RAID 1 0 8 HP Smart Array P410i No Cache HP BladeSystem c7000 Enclosure 1 x HP Onboard Administrator Module 1 x 1Gb Ethernet Pass Thru Module for c Class BladeSystem 6 x 2450W
5. EKIN eV SPECpowWer ssj2008 Staridard deett ege eerie s une ox eve buie ene vut dede Re eie RIEN IU SNR e vn vga sates BIOS Settings iioc eode tra da TEE ERR Dre Ua e Eder ina pO o Eu Lea Beg TR EEE Er Geeta Gib hie ovate dere bint d ao deed ed Fat he s QOS HUI E E SPECpower ssjZ008 Configuration 135i is iresi en CEEE EEEE E CEEE epi pun e Or M epp Pe puh se Ori dps er uh sea eds Appendix B Blade Server Hardware Configuration ccccsccscescesccsccsccsccsccescsscescescescescessessesees Appendix C Server Firmware and Drivers ceeeseeeeeeeeeeee eese eee eene ehh e hene ree tree hr eren Poor edpeiilp ime e i Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Table 1 Detailed Blade Solution Configurations ccccccscccccccceesscccccecenessscseceseeesssseees 9 Table 2 Dell Blade Solution Cost Breakdown ccsscccceccesscccccccessssscceccsesessssesesseeenenes 16 Table 3 HP Blade Solution Cost Breakdown 2 0455 9d Soa ees Er nue Eno a e RS IR Y TERR a REFED vA RE iuiedus 16 Table 4 Blade Server Hardware Configuration cccccsesssccscccceesssscccccecessssscseseeeseeees 22 Table 5 Server Firmware and Drivers iure entra nr e i hr en ae in p rn Re P Eti o d 23 Figure 1 Blade Solution Comparison Normalized Values cessere 10 Figure 2 Blade Solution Power Efficiency sci sii erui e er sur e epe reatu Ene ARR PER REOR
6. INHE ERREUR UD NEA 11 Figure 3 Blade Solution Full Chassis Performance 12s veux EN Rie AI RUE EARN ER GUN LESE DN SR RN sien 12 Figure 4 Blade Solution Power Consumption at 100 Target Load eeeeeeeee eee 13 Figure 5 Blade Solution Power Consumption at Idle eese eere eene een h eher eonun 14 Figure 6 Price as Tested cueste tus d Eos Parada ad dtc Ge ascatal fa RN ala E OT fcu set 15 Figure 7 Blade Solution Performance per Dollar oa etd kir ra ERE eI d aora a LEX Sa o PRA Ern no dex oS 17 Figure 8 Dell Blade Solution Benchmark Results sccccsesesccccccccesssssccccscesessscceseeees 24 Figure 9 HP Blade Solution Benchmark Results ceeceeeeeeeeeeeee eee eee enne nnn 25 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Executive Summary Introduction Dell commissioned its Solutions Performance Analysis team to compare the power efficiency of full chassis blade solutions using 4 socket blades from Dell and Hewlett Packard HP In July 2011 the Dell PowerEdge M915 blade server achieved the highest SPECpower ssj2008 score of any 4 socket server on the market so eight of these blades and a PowerEdge M1000e enclosure were used as the PowerEdge solution For the ProLiant solution HP s similar BL685c G7 could have been chosen for this study However the BL680c G7 is HP s top selling 4 socket blade so it was chosen to al
7. M tunings for SPECpower ssj2008 for the IBM J9 JVM when running with larger memory configurations Xms1875m Xmx1875m Xmnl1400m Xaggressiv Xcompressedrefs Xgcpolicy gencon XlockReservation Xnloa XtlhPrefetch Xlp On the PowerEdge M915 blades the following bindings were used to ensure that each of the eight JVMs ran on six logical processors start affinity 3F FCO 3F000 FC0000 3F000000 FC0000000 3F000000000 FC0000000000 15 JVM build 2 4 J2RE 1 6 0 IBM J9 2 4 Windows Server 2008 amd64 64 jvmwa64 60sr5 20090519 35743 20 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions On the ProLiant BL680c G7 blades the following bindings were used to ensure that each of the 16 JVMs ran on 4 logical processors start affinity F F0 F00 F000 F0000 F00000 F000000 F0000000 F00000000 F000000000 F0000000000 F00000000000 F000000000000 F0000000000000 F00000000000000 F000000000000000 Power Meter Configuration Power consumption for each blade chassis under test was monitored by two Yokogawa WT210 Digital Power Meters Both chassis under test had a total of 6 PSUs so each Yokogawa WT210 monitored power for 3 PSUs 21 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Appendix B Blade Server Hardware Configuration Table 4 Blade Server Hardware Configuration Memory Modules Total RAM in each blade GB Vendor an
8. MHz from the default of 1333MHz and changed their HyperTransport frequency to HT1 Virtualization was not used in these tests so AMD Virtualization support was disabled on both servers For the Dell PowerEdge M915 blades the following settings were used e DRAM Prefetcher Disabled e Hardware Prefetch Training on Software Prefetch Disabled e Hardware Prefetcher Disabled e HT mode set to HT1 e Memory Speed set to 1066MHz in BIOS e AMD Virtualization was disabled Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions For the HP ProLiant BL680c G7 blades the following settings were used e Disabled Intel Turbo Boost Technology e Dynamic Power Savings Mode Response set to Slow e Enabled DIMM Idle Power Saving Mode e Disabled HP NC553i 10Gb 2 Port FlexFabric Adapter 2 in BIOS e Disabled HP NC553i 10Gb 2 Port FlexFabric Adapter 3 in BIOS e Disabled Intel SATA Controller 1 in BIOS e Disabled Hardware Prefetcher in BIOS e Disabled Adjacent Sector Prefetch in BIOS e The HP ProLiant BL680c G7 blades provide an option called Collaborative Power Control in its ROM Based Setup Utility RBSU which is Enabled by default Disabling this option on all blades did not improve overall SPECpower_ssj2008 results and resulted in Invalid runs so this setting was left at the default Enabled setting for the runs which produced the data used in the comparison OS Tuning To improve Java perf
9. Platinum Rated In order to compare the two solutions as closely as possible each blade server had the same amount of system memory and each blade was fully populated with memory the M915 blades each had 32 4GB RDIMMs for a total of 128GB system memory and the BL680c G7 blades each had 64 2GB RDIMMs also for a total of 128GB system memory This configuration also meant that the total number of DIMMs across all blades was 256 per solution Dell only sells low power LV DIMMs in the PowerEdge M915 blades so 4GB LV DIMMs were used for the comparison The HP ProLiant BL680c G7 blades only offer LV DIMMs on their Intel Xeon E7 4800 series On the HP Smart Array P410i RAID 1 0 is the only option available in a 2 hard drive configuration Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions models which were not available at the time the solutions were purchased so standard voltage 2GB RDIMMs from HP were used for the HP blade solution Results The Dell blade solution with PowerEdge M915 blades delivered 8796 higher raw performance than the HP solution with ProLiant BL680c G7 blades The Dell solution also has 9296 higher power efficiency and even with 8 blades compared to 4 blades for the HP solution consumes 35 less power at idle Figure 1 Blade Solution Comparison Normalized Values Normalized Blade Solution Comparison Power Efficiency Performance and Power Draw Power Efficie
10. Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions A Dell Technical White Paper Brian Bassett and Chris Christian Solutions Performance Analysis Dell Enterprise Solutions Group Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions THIS WHITE PAPER IS FOR INFORMATIONAL PURPOSES ONLY AND MAY CONTAIN TYPOGRAPHICAL ERRORS AND TECHNICAL INACCURACIES THE CONTENT IS PROVIDED AS IS WITHOUT EXPRESS OR IMPLIED WARRANTIES OF ANY KIND 2011 Dell Inc All rights reserved Reproduction of this material in any manner whatsoever without the express written permission of Dell Inc is strictly forbidden For more information contact Dell Dell the DELL logo and the DELL badge and PowerEdge are trademarks of Dell Inc Microsoft Windows and Windows Server are either trademarks or registered trademarks of Microsoft Corporation in the United States and or other countries Intel and Xeon are registered trademarks of Intel Corporation in the U S and or other countries SPEC and the benchmark names SPECpower ssj and SPECjbb are trademarks of the Standard Performance Evaluation Corporation Other trademarks and trade names may be used in this document to refer to either the entities claiming the marks and names or their products Dell Inc disclaims any proprietary interest in trademarks and trade names other than its own August 2011 Revision 1 0
11. Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Appendix D Results The full reports from the SPECpower ssj2008 runs used in this comparison are attached to this whitepaper for reference The first page of each is shown here for convenience Figure 8 Dell Blade Solution Benchmark Results SPECpower ssj2008 Dell Inc 3 Multi Node Set sut WARNING For point 8 elapsed nanoTime 240108845676 ms elapsed currentTimeMillis 240007 ms Set sut WARNING For point 3 elapsed nanoTime 240015647518 ms elapsed currentTimeMillis 240131 ms Set sut WARNING For point 1 elapsed nanoTime 240250333681 ms elapsed currentTimeMillis 240132 ms Set sut WARNING For point 2 elapsed nanoTime 240125634304 ms elapsed currentTimeMillis 240287 ms Set sut WARNING For point 0 elapsed nanoTime 240093929556 ms elapsed currentTimeMillis 240241 ms Set sut WARNING For point 0 elapsed nanoTime 240060738204 ms elapsed currentTimeMillis 240178 ms Set sut WARNING For point 1 elapsed nanoTime 240030140492 ms elapsed currentTimeMillis 240131 ms Benchmark Results Summary Performance Power Performance to Power Ratio Average Performance to o 500 1 000 1 500 2 000 2 500 3 000 2 425 overall ssj ops watt Target Actual 70 096 59 9 Target Load 8 E gt o 1 000 2 000 3 000 4 000 Average Active Power W 24 Power Efficiency Comparison of the Dell Powe
12. VM HP ProLiant BL685c G7 4 chips 48 cores 48 threads 1 832 929 SPECjbb2005 bops 8 JVMs 229 116 SPECjbb2005 bops JVM Based on best SPECjbb2005 results published on http www spec org as of July 7 2011 For latest SPECjbb2005 benchmark results visit www spec org osg jbb2005 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Appendix A Test Methodology SPECpower_ssj2008 Standard SPECpower_ssj2008 is an industry standard benchmark created by the Standard Performance Evaluation Corporation SPEC to measure a server s power and performance across multiple utilization levels SPECpower_ssj2008 consists of a Server Side Java SSJ workload along with data collection and control services SPECpower_ssj2008 results portray the server s performance in ssj_ops server side Java operations per second divided by the power used in watts ssj_ops watt SPEC created SPEcpower_ssj2008 for those who want to accurately measure the power consumption of their server in relation to the performance that the server is capable of achieving with ssj2008 workload SPECpower_ssj2008 consists of three main software components e Server Side Java SSJ Workload Java database that stresses the processors caches and memory of the system as well as software elements such as OS elements and the Java implementation chosen to run the benchmark e Power and Temperature Daemon PTDaemon Program that controls
13. and reports the power analyzer and temperature sensor data e Control and Collect System CCS Java program that coordinates the collection of all the data For more information on how SPECpower_ssj008 works see http www spec org power_ssj2008 All results discussed in this whitepaper are from compliant runs in SPEC terminology which means that although they have not been submitted to SPEC for review Dell is allowed to disclose them for the purpose of this study All configuration details required to reproduce these results are listed in Appendices A B and C and all result files from the runs compared are included in Appendix D Both servers were configured by installing a fresh copy of Microsoft Windows Server 2008 Enterprise R2 Service Pack 1 with the operating system installed on a two hard drive RAID 1 choosing the full installation option for each The latest driver and firmware update packages available to both servers were installed at the beginning of this study Refer to Appendix B for details The Dell System Performance Analysis Team ran SPECpower_ssj2008 ten times per configuration across both servers and chose the run with the highest overall ssj_ops watt for each configuration to compare BIOS Settings BIOS settings differed between the two manufacturers so we tuned for best known SPECpower_ssj2008 performance results To improve power efficiency we changed the memory speed of the M915 blades to 1066
14. ar spent on the solution compared to just 54 ssj ops at 100 target load for each dollar spent on the HP solution as seen in Figure 7 Figure 7 Blade Solution Performance per Dollar E Dell Blade Solution HP Blade Solution 12 The 8 node M915 achieved 14 982 232 ssj_ops 100 target load and a SPECpower ssj2008 result of 2 425 overall ssj_ops Watt compared to the 4 node BL680c G7 with 7 996 829 ssj_ops 100 target load and a SPECpower ssj2008 result of 1 260 overall ssj ops Watt Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Industry Leading 4 Socket Server Power Efficiency The 8 node Dell M915 blade solution in this comparison achieved an overall SPEC score of 2 425 even when configured as it would in a typical datacenter with redundant hard drives and all DIMM slots filled With minor changes to the configuration using hardware which can be ordered standard from www dell com the 8 blade M915 solution achieved an overall SPECpower score of 2 716 which is the highest score of any 4 socket rack or blade server published at www spec org as of July 2011 Future Platform Updates The PowerEdge M915 blades in the Dell solution were equipped with AMD Opteron model 6176 processors the most power efficient processors available for the M915 blades at the time equipment for this study was purchased Likewise the ProLiant BL680c G7 blades in the HP solution were equipped with Intel Xeo
15. d model number Type Speed MHz Speed in system as tested Timing latency Number of RAM modules Rank organization Hard Disk Vendor and model number Number of disks in system Size GB RPM Type RAID Type Controller Operating System Name Build number File system Language Network Adapter Vendor and model number Type Dell PowerEdge M915 128 Samsung M393B523CHO YH9 PC3L 10600R 1333 1066 CASTS 32 x 4GB Dual Rank Dell P N 0J084N 2 146 15 000 SAS 6 Gbps RAID 1 Dell PERC H200 Microsoft Windows Server 2008 R2 Enterprise SP1 7601 NTFS English 2 x Broadcom BCM57712 k Dual Port 10Gb Integrated HP ProLiant BL680c G7 128 Samsung M393B5673FH0 CH9Q5 PC3 10600R 1333 1066 CAS 9 64 x 2 GB Dual Rank HP 518216 002 2 146 15 000 SAS 6 Gbps RAID 1 0 HP Smart Array P410i Microsoft Windows Server 2008 R2 Enterprise SP1 7601 NTFS English 6 HP NC553i 10Gb FlexFabric adapter ports Integrated 16 On the HP Smart Array P410i RAID 1 0 is the only option available in a 2 hard drive configuration 22 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Appendix C Server Firmware and Drivers Table 5 Server Firmware and Drivers Driver Firmware Dell PowerEdge M915 HP ProLiant BL680c Versions G7 Network Firmware 6 2 16 31022517 7 HBA Firmware 07 02 42 00 B5 Management Controller 3 3 0 0 Driver 23
16. ization than the chassis full of 4 HP ProLiant BL680c G7 blades Power e Despite having 87 greater performance the Dell solution with 8 PowerEdge M915 blades consumes only 14 more power at 100 utilization compared to the HP solution with just 4 ProLiant BL680c G7 blades e Despite having twice as many servers the Dell blade solution consumes 35 less power at idle than the HP blade solution In the published result the 8 node M915 achieved 14 793 524 ssj_ops 100 target load and a SPECpower ssj2008 result of 2 716 overall ssj_ops Watt Benchmark results based on results published at www spec org as of July 2011 For the latest SPECpower_ssj2008 benchmark results visit http www spec org power ssj2008 results power ssj2008 html 2 According to IDC Q1 2011 Server Tracker May 22 2011 http h10010 www1 hp com wwpc us en sm WF05a 3709945 3709945 3328410 241641 3722793 4268690 html Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Price e Priced as configured the Dell PowerEdge solution including 8 M915 blades and the M1000e Blade Chassis costs 21 less overall than the HP ProLiant solution which includes 4 BL680c G7 blades and the BladeSystem C7000 enclosure e The Dell solution s lower cost and greater power efficiency lead to a 59 better price performance watt ratio Rack Density e The Dell solution provides 8 servers per 10U chassis compared to the HP solutio
17. losure 1 9 701 00 9 701 00 Dell Total Cost 116 405 00 Percent less cost for the Dell Solution 21 SPECpower_ssj2008 Overall ssj_ops watt 2 425 Dell Solution ssj_ops watt 48 00 Dell Percent less ssj ops watt 5996 Table 3 HP Blade Solution Cost Breakdown HP Blade Solution Qty Each Total ProLiant BL680c G7 blades 4 34 495 00 137 980 00 BladeSystem c7000 Enclosure 1 10 123 00 10 123 00 HP Total Cost 148 103 00 SPECpower_ssj2008 Overall ssj ops watt 1 260 HP Solution ssj ops watt 117 54 The lower cost and higher power efficiency of the Dell solution lead to a cost of just 48 00 for each point of overall SPECpower ssj2008 score 59 less than the 117 54 per point of overall SPECpower_ssj2008 score with the HP solution M915 and M1000e prices in U S dollars from Dell com June 14 2011 1 BL 680c G7 and C7000 prices in U S dollars from HP authorized reseller June 14 2011 11 The 8 node M915 achieved 14 982 232 ssj_ops 100 target load and a SPECpower ssj2008 result of 2 425 overall ssj_ops Watt compared to the 4 node BL680c G7 with 7 996 829 ssj_ops 100 target load and a SPECpower_ssj2008 result of 1 260 overall ssj_ops Watt Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Performance per Dollar The lower price and higher overall performance of the Dell solution allow it to provide 129 ssj ops at 100 target load for each U S doll
18. low examination of HP s claims of that server s unique 4P blade performance and improved power efficiency Due to its full height double wide form factor only four of these blades can fit in a 10U BladeSystem c7000 blade enclosure Using the industry standard SPECpower_ssj2008 benchmark the performance power draw and performance watt of full chassis configurations of both blade solutions were tested The blade servers were compared configured as similarly as possible given their architectural differences with identical memory and hard drive selections and best known CPU choices for highest performance watt ratio To ensure the lowest power draw both blade chassis were equipped with six of their Platinum rated power supplies for all tests The results showed the Dell solution using 8 PowerEdge M915 costs less than the HP solution consisting of just 4 BL680c G7 blades the maximum that can fit in the BladeSystem c7000 enclosure The Dell solution also consumed less power at idle provided higher raw performance and achieved a better performance watt ratio The Dell solution also fits twice as many servers in a 10U blade enclosure Key Findings Performance Watt e The Dell solution with 8 PowerEdge M915 blades achieved a 92 higher performance watt ratio across all load levels than the HP solution using 4 ProLiant BL680c G7 blades Performance e The chassis full of 8 PowerEdge M915 blades achieved 87 higher raw performance at 100 util
19. n X7560 processors the most power efficient processors available for that platform at the time equipment for the testing was purchased When AMD Opteron 6200 series processors are available for the Dell solution and Intel Xeon E7 series processors are available for the HP solution Dell plans to update this study to show what changes if any these next generation processors bring to the relative performance of the two blade solutions Summary The focus of this study was to examine the overall performance and power efficiency of full chassis solutions utilizing the latest blades released by Dell M915 and HP BL680c G7 The Dell solution with eight PowerEdge M915 blades was shown to provide 92 greater overall performance with 87 greater power efficiency for 21 less cost than the HP solution which could only accommodate four BL680c G7 blades An HP solution comprised of eight BL685c G7 blades could also have been used for the comparison Given the PowerEdge M915 blade s 6 performance advantage over the BL685c G7 in the SPECjbb2005 benchmark the Dell PowerEdge M915 blade solution would be expected to have an overall performance advantage in such a comparison 13 Based on results for the 8 node M915 as published at www spec org as of 15 July 2011 For more information about SPECpower see www spec org power_ssj2008 14 Source Dell PowerEdge M915 4 chips 48 cores 48 threads 1 949 160 SPECjbb2005 bops 8 JVMs 243 645 SPECjbb2005 bops J
20. n which provides only 4 servers per 10U chassis Test methodology and detailed results are documented in this paper The comparison presented here is based on the respective enterprise class servers configured as similarly as possible and currently shipping by Dell and HP Results based on SPECpower_ssj2008 performance testing by Dell Labs in June 2011 For the latest SPECpower ssj2008 benchmark results visit http www spec org power ssj2008 results power ssj2008 html For latest SPECjbb2005 benchmark results visit http www spec org osg jbb2005 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions Methodology SPECpower ssj2008 is an industry standard benchmark created by the Standard Performance Evaluation Corporation SPEC to measure a server s power and performance across multiple utilization levels Appendix A details the test methodology used by Dell Appendices B and C provide detailed configurations for the tests and Appendix D provides detailed report data that supports the results in this paper Full disclosure reports from the valid SPECpower ssj2008 runs used in this comparison are attached to this whitepaper for reference Configurations The blade servers in both solutions were configured for their best known SPECpower_ssj2008 configurations and were matched as closely as possible given the differences between the architectures Both solutions used the maximum number of
21. ncy Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions The Dell solution s greater power efficiency is illustrated in Figure 2 The SPECpower ssj2008 overall ssj ops watt for the 8 node Dell solution is 2 425 92 higher than the 4 node HP solution s 1 260 ssj ops watt Figure 2 Blade Solution Power Efficiency Blade Solution Power Efficiency Performance watt higher is better 2 425 Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions The PowerEdge M915 blade solution s density advantage is illustrated in Figure 3 While both solutions consume 10U of rack space the Dell solution can accommodate 8 M915 blades compared to only 4 BL680c G7 blades in the HP solution This leads to the Dell solution s 87 advantage in aggregate full chassis performance Figure 3 Blade Solution Full Chassis Performance Blade Solution Full Chassis Performance Higher is better 18 000 000 e 16 000 000 14 982 232 p 7 996 829 The 8 node M915 achieved 14 982 232 ssj_ops 100 target load and a SPECpower ssj2008 result of 2 425 overall ssj_ops Watt compared to the 4 node BL680c G7 with 7 996 829 ssj_ops 100 target load and a SPECpower ssj2008 result of 1 260 overall ssj ops Watt Power Efficiency Comparison of the Dell PowerEdge M915 and HP ProLiant BL680c G7 Blade Solutions SPECpower ssj2008 includes a measurement of power while the servers are at 100
22. ormance large pages were enabled by entering Control Panel Administrative Tools gt Local Security Policy gt Local Policies gt User Rights Assignment gt Lock Pages in Memory An option was changed to add Administrator Operating System Power Management mode for both solutions was left at Balanced the default and Balanced mode was edited to turn off the Hard Drive after 1 minute On the HP ProLiant BL680c G7 blades the Minimum Processor State was changed to 0 Some platforms produce better overall SPECpower_ssj2008 scores using the Windows 2008 Power option Power saver with Maximum Processor set to 100 but on both blade solutions this did not significantly change the overall score so this setting was left at the default Balanced setting for the runs which produced the data used in the comparison We configured all servers with a separate IP address on the same subnet as our SPECpower ssj2008 controller system where the Director CCS and PTDaemon components were located and connected all servers and the controller system to an external PowerConnect Gigabit Ethernet switch through NIC 1 for their respective runs SPECpower ssj2008 Configuration IBM J9 Java Virtual Machine JVM was used for both solutions as this JVM provided the best performance for SPECpower_ssj2008 of any of the available choices at the time that this study was undertaken The following JVM options were used on both servers as they are the best known JV
23. rEdge M915 and HP ProLiant BL680c G7 Blade Solutions Figure 9 HP Blade Solution Benchmark Results SPECpower ssj2008 Jun 7 2011 Unpublished Set sut WARNING For point 3 elapsed nanoTime 2400263851 14 ms elapsed currentTimeMillis 240241 ms Set sut WARNING For point 1 elapsed nanoTime 240200467528 ms elapsed currentTimeMillis 240349 ms Set sut WARNING For point 0 elapsed nanoTime 240054160612 ms elapsed currentTimeMillis 240178 ms Set sut WARNING For point 1 elapsed nanoTime 240133818681 ms elapsed currentTimeMillis 240349 ms Set sut WARNING For point 3 elapsed nanoTime 240027992607 ms elapsed currentTimeMillis 240178 ms Set sut WARNING For point 1 elapsed nanoTime 240020123996 ms elapsed currentTimeMillis 240225 ms Set sut WARNING For point 0 elapsed nanoTime 240205610034 ms elapsed currentTimeMillis 240412 ms Set sut WARNING For point 3 elapsed nanoTime 240021225761 ms elapsed currentTimeMillis 240225 ms Benchmark Results Summary Performance Power Performance to Power Ratio Average Performance to Target Load 30 _30 1 2 421 491 2 848 850 109 101 Emme rs aa i 2 291 Average Active Power W 1 000 2 000 3 000 4 000 25
Download Pdf Manuals
Related Search