HP ProLiant DL585 Dual-Core server benchmark results for Microsoft Exchange Server 2003

Informations

• Publié par: Pidel
• Nombre de lectures: 46
• Langue: Slovak

Extrait

 

 

 

HP ProLiant DL585 Dual-Core server benchmark results for Microsoft Exchange Server 2003

  Executive summary............................................................................................................................... 2   ProLiant DL585  the server behind the outstanding performance ............................................................. 3   Overview ........................................................................................................................................ 3   ProLiant DL585 server features........................................................................................................... 3   HP StorageWorks 5000 Enterprise Virtual Array ..................................................................................... 4   Description......................................................................................................................................4   Features..........................................................................................................................................4   Test methodology and workload ............................................................................................................ 5   Test result highlights ............................................................................................................................. 7   Appendix A  Detailed test results.......................................................................................................... 8   Descriptive terms..............................................................................................................................8   Test results .......................................................................................................................................... 9   Response times .............................................................................................................................. 16   Message throughput ....................................................................................................................... 16   Test configuration .............................................................................................................................. 17   Load generator configuration........................................................................................................... 19   Additional configuration and tuning ..................................................................................................... 20   Appendix B  LoadSim changes from previous version ........................................................................... 21   Appendix C  MMB2 and MMB3 workload comparison ........................................................................ 22   For more information.......................................................................................................................... 30   HP ActiveAnswers ...................................................................................................................... 30   HP Microsoft Exchange Solutions.................................................................................................. 30    

 

 

Executive summary

In December 2005, HP produced new Microsoft® Exchange MAPI Messaging Benchmark (MMB3) results using Microsoft Windows® Server 2003 and Microsoft Exchange Server 2003, on an HP ProLiant DL585 server powered by four Dual-Core AMD Opteron Model 880 processors (2.4 GHz, 1 MB L2). The MMB3 benchmarking workload and methodology serves as the standard for Exchange Server 2003 MAPI server comparison. The MMB3 workload is characteristic of a medium corporate mail environment. Using the Microsoft LoadSim utility, the ProLiant DL585 server was tested at the HP Performance Center in Nashua, New Hampshire. HP achieved world-class Microsoft Exchange Server 2003 scalability results of 12,000 MMB3 on a ProLiant DL585 server equipped with: •   Four Dual-Core AMD Opteron Model 880 2.4GHz - 1MB L2 processors •   320  36.4 GB 15K hard drives for the database files •   56  36.4 GB 15K hard drives for transaction log files •   2  36.4 GB 15K hard disk drives for the Exchange files, operating systems, and Active Directory files

The quad processor ProLiant DL585 server achieved the following: •   Average CPU utilization rate of55.9%during the12,000MMB3 test •   Weighted 95th percentile response-time score was279milliseconds (ms) •   Average send-queue size for the four-hour steady-state period was81messages

2 

 

ProLiant DL585  the server behind the outstanding performance Overview The HP ProLiant DL585 server delivers the highest 32-bit performance of any 4-processor server for Microsoft Exchange Server, as well as offering 64-bit capabilities, enabling customers to deploy a common 4-processor platform for their high performance 32-bit and future 64-bit computing requirements. The combination of Dual-Core Opteron processors, large memory footprint and 1GHz HyperTransport has produced the fastest x86 server in the industry . The DL585 combines AMD Opteron processor technology, best-in-class management and outstanding uptime features in an enterprise-class system ideal for large data center deployments. The 4-processor HP ProLiant DL585 now supports 2.4GHz dual-core processors, 2.8 single-core processors, 1GHz HyperTransport, and PC2700 and PC3200 memory, producing a server ideal for space constrained customers looking for leading performance and management in a rack-optimized form factor.

  Figure 1.HP ProLiant DL585 server

 

  ProLiant DL585 server features •   Dual-Core AMD Opteron Model 880 (2.4GHz-1MB L2) Dual-Core AMD Opteron Model 875 (2.2GHz-1MB L2) Dual-Core AMD Opteron Model 865 (1.8GHz-1MB L2) Model 854 (2.8GHz-1MB L2 single-core) Model 852 (2.6GHz-1MB L2 single-core) Model 850 (2.4GHz-1MB L2 single-core) Model 848 (2.2GHz-1MB L2 single-core) •   PC2700 DDR running at 266MHz, 48GB ofUp to 128GB of dual channel PC2700 running at 333MHz or 32MB of PC3200 running at 400MHz •   Smart Array 5i Plus controller, with 64MB memory and battery-backed write cache •   HP embedded NC7782 Dual-Port PCI-X Gigabit Server Adapter 3  

 

•8 I/O slots: 6 x 64-bit/100 MHz PCI-X expansion slots and 2 x 64-bit/133MHz    PCI-X slots •   HyperTransport (processor interconnect) runs at 800MHz with 8.0GB of bandwidth •   On-die memory controller which runs at processor core frequency

  HP StorageWorks 5000 Enterprise Virtual Array Description The HP StorageWorks 5000 Enterprise Virtual Array (EVA5000) is a high performance, high capacity and high availability "virtual" RAID storage solution for the high-end enterprise class marketplace that removes the time, space and cost boundaries of traditionally architected storage. The EVA5000 is designed for the data center where there is a critical need for improved storage utilization and scalability while meeting application specific demands for consistently high transaction I/O and MB data rate performance, seamless capacity expansion, instantaneous replication, and simplified storage administration. The EVA5000 provides an enhanced service portfolio combining hardware and services into a single part number to rapidly deploy the EVA and maintain a solid storage platform. HP offers a full spectrum of complementary EVA hardware and software product service solutions ranging from the new Foundation Service Solution, the base level of service included with every EVA, to Critical Service, the HP support for mission critical environments.

  Figure 2:StorageWorks EVA5000

 

  Features •   Management: allows management of large SAN configurations, reduces management time and lowers costs • Performance: outstanding self-tuning performance accomplishes more in less time, scales   capacity on-demand and minimizes data administration 4  

 

•   Unique virtualization features: state-of-the-art architecture improves performance, optimizes disk usage, and allows for easy dynamic storage expansion •   Virtually instantaneous Snapclone: allows immediate use of the clone copy, saving significant time •   Virtually capacity-free Snapshot (Vsnap): saves customers disk space and improves disk utilization efficiency during local replication operations •   High density: Supports up to 35 TB of versatile storage capacity with any combination of high performance and low cost FATA drives using existing EVA enclosures •   Support: enables effectiveness of deployment and delivers the proactive services required to help maintain a stable, highly available EVA storage platform over time •   Remote service tools: ISEE , the remote service tool for EVA, provides, at no additional cost, event monitoring and proactive diagnostic capability via transmission of EVA status to the HP service center for proactive problem solving Test methodology and workload

For Microsoft Exchange Server 2003, the benchmarks were measured using the MAPI Messaging Benchmark 3 (MMB3). The MMB3 workload, for LoadSim 2003, is a modification of the previous MMB2 workload. It is designed to include new features from Microsoft Exchange Server 2003 and Microsoft Outlook 2003. This workload achieves the following: •   Uses the Microsoft Outlook 2003 client •   Introduces Smart Folders •   Introduces the use of server-side rules •   Allows the message distribution to be composed of a larger message size than MMB2 •   Increases the mailbox size to 100MB per user •   Removes journaling from the benchmark For more detailed information regarding the differences between MMB2 and MMB3, please refer to Appendix C  MMB2 and MMB3 workload comparison in this document. This test measures the messaging throughput of a single server, single site topology. Its purpose is to measure the maximum throughput of a Microsoft Exchange Server on this hardware configuration.

Note:This test can provide a benchmark for comparing hardware and/or software products, but cannot be used as a deployment guide for production environments. For deployment specific information, visit tp://wwwhtulosnoit.ph./mocgeexs/anch .

The MMB3 benchmark does not account for: •   Usage profiles not matching that of the LoadSim MMB3 profile

5 

 

  • •   •   •   •   •     • •     •  

Per user storage and per server backup requirements Fault tolerance requirements, such as protected storage (RAID 0+1, RAID 5) for the system/page file volume, information store and transaction logs Anti-virus and management processes and effects on the server UBE/UCE (spam) mail flow Workloads other than MAPI private folder access. This includes Public Folder, NNTP, POP3 and other email interfaces Multiple Exchange Server deployments, where additional resources are required to forward mail intra-site Connectors, Links and Replication to remote Exchange sites Network topologies, bandwidth availability, latency requirement and SLA related factors like QOS (Quality of Service) and fail-over path issues Manageable database sizes and partitioning beyond the 2 Storage Group, 2 database/SG configuration

6 

 

Test result highlights

Hewlett Packard Server: Test results MMB3 score: Response time: CPU utilization: Avg. queue: Msgs. Submitted Msgs. Delivered Msgs. Sent Server configuration CPU: CPU count: RAM: Secondary cache: Operating system: Storage:

Controller: NIC:    

HP ProLiant DL585

12,000 279 milliseconds (ms) 55.9% 81 517,757(4-hour steady state period)   1,286,202(4-hour steady state period) 517,709(4-hour steady state period)  

Dual-Core AMD Opteron 2.4 GHz / 1 GHz HyperTransport 4 Physical 8 GB 1 MB L2 Cache per core Microsoft Windows Server 2003 Enterprise Edition 320  36.4GB 15K  Information Store Databases (EVA5000)  56  36.4GB 15K  Transaction Log files (MSA1500)  2  36.4GB 15K  Operating system, Exchange files and  Active Directory Integrated HP Smart Array 5i Plus controller Dual Port Fibre Channel Card (2Gb) Dual Port NC7782

7 

 

Appendix A  Detailed test results

Descriptive terms

Messages Submitted  Submitted calls made by clients. This equates to total messages sent by users. Messages Sent  Messages that the Store sends to the categorizer in Inetinfo (SMTP Service in particular).

Note:All messages  even MAPI messages  are sent to the categorizer, as this replaces the MTA for all but communication via X.400, with a Microsoft Exchange 5.5 server.

Message Recipients Delivered  Separate mailboxes where messages have been delivered. Message Opens/Sec  Messages accessed for reading per second. Folder Opens/Sec  Folders opened for browsing per second. RPC Read Bytes/Sec  Bytes read from clients, sent via RPCs. RPC Write Bytes/Sec  Bytes written to clients, sent via RPCs. IS Send Queue Average Length  Send Queue Size is the number of messages in the private information store's send queue.

8 

 deutxenoc 1nitnTa (e bl    

9 

12,000MMB3s MAPI Messaging Benchmark 3 (MMB3) Exchange MAPI 4 Hours 8 Hours   517,757 1,286,202 517,709 2.48 517,757 129,439 321,551 129,427 346,737 6,258,717

 page)     ces / setyB daReC RP ) ndcoser   s cese/ B tyritePC WR saess gentSe h / oh / ru M saction Load (pe uo r rTna ho/ d te uregasseM   timbuS s s Deientred liveseasMcepiegR  ontiacnsraT)ruoh rep( daoL  ssagl MeTota ttdebuimseS  / d ssMeliDereve ttim de sega buS  Mes Sent  pieitn sasegR ce  derevi  MalotTs gesaes oTatssga leMcipie Re Delents T  la latoseM egasSus itbmd te se t Tarnsactions in Tot foT tg hLne  tatedy SStea fo htgneL l cotoroP elif kramorP Bchen markenchd  LoaS   pu amyr deB optr es rstTe um   S tsulT  .elifo  :1 elba  for the MMB3 prifactnylg ertareueeqs st sisniigetar fo eilcr tn essaPI Mw MAe ne  s hT ustl terT   se esurasme) B3MM (3 kramhcneB gniga speciferms of up tnit t rhuohg , nsecex aeriocto elsu fp  ci iforowkruo rad.yni g oveuted 8 hr an effid siorf tnerbes hi   Tk arhmncgnt ah taw ssudem the MMB2 setti 000 t  ni  tah eht it wExh anch 2ge

 

  Test results (continued) Processor % Processor Time Database Database cache size Table opens/sec Memory Utilization Available Bytes Cache Faults/sec Free System Page Table Entries Pages / sec Pool Non-paged Bytes (Bytes) Pool Paged Bytes (Bytes) System Cache Resident Bytes Transition Faults/sec MSExchangeIS Mailbox Folder Opens / sec Message Opens / sec MSExchangeIS Send Queue Average Length MSExchangeIS Receive Queue Average Length (Table 1 continued next page)  

Average Max Min 49.39 61.50 0.00 Average Max Min 1156756753.07 1174405120.00 442671104.00 1641.80 2205.48 0.00 Average Max Min 5857366835.20 7899791360.00 5629833216.00 1318.86 2164.75 0.00 16711.33 17957.00 16397.00 3.49 24.23 0.00 29538116.27 30605312.00 22990848.00 39584955.73 41742336.00 23830528.00 119472025.60 137113600.00 106635264.00 22.66 1129.01 0.00 Average Max Min 49.82 228.87 0.00 106.06 164.20 0.00 74.85 184.00 0.00 0.00 0.00 0.00

10 

 

Test results (continued) MSExchangeIS  Active User Count RPC Average Latency (ms) RPC Num. of Slow Packets RPC Packets/sec RPC Read bytes/sec RPC Requests RPC Operations/sec RPC Write bytes/sec TempTable Current MSExchangeIS VM Largest Block Size MSExchangeIS VM Total 16MB Free Blocks MSExchangeIS VM Total Free Blocks MSExchangeIS VM Large Free Block Bytes Paging File % Usage (_Total) Processor Utilization System Processor Utilization (%)   System Processor Interrupts/sec (Total)   Process % CPU Time  Store Process % CPU Time  Inetinfo Exchange server is also domain controller? (yes/no) Process % CPU Time  LSASS (on domain controller) (Table 1 continued next page)

Average Max Min 192.43 343.00 0.00 14.44 36.00 0.00 2.51 10.00 0.00 1439.37 1835.92 0.00 351520.89 552081.55 0.00 20.45 35.00 0.00 2397.25 3197.62 0.00 6100943.11 8605486.09 0.00 7.65 18.00 0.00 694054502.40 1029259264.00 594542592.00 3.30 14.00 2.00 265.75 287.00 177.00 772253286.40 2082119680.00 622829568.00 Average Max Min 7.36 30.30 6.87 Average Max Min 49.39 61.50 0.00 6485.43 7379.76 0.00 344.72 428.18 0.00 8.32 11.69 0.00 Y Y Y 8.74 18.18 0.00

11