SPECweb96 Result =============================================================================== || Dell Computer Corp. : PowerEdge 8450/550 || 22007 SPECweb96 Microsoft : IIS 5.0 and SWC 2.0 || || =============================================================================== Throughput Response ops/sec msec 2310 4.8 4620 5.2 6931 5.8 9241 6.4 11551 7.1 13861 8.1 16173 9.3 18483 11.5 20683 13.1 22007 15.0 =============================================================================== General Availability Dates Hardware Sep-99 Operating System Oct-99 HTTP Software Dec-99 System Hardware Vendor Dell Computer Corp. Model PowerEdge 8450/550 Processor 550MHz Pentium III Xeon # of Processors 8 Primary Cache 16KBI+16KBD on chip Secondary Cache 2MB(I+D) off chip Other Cache None Memory 4GB Disk Controllers Symbios Ultra-SCSI, Dell PERC2/SC Disk Subsystem 9 9GB 10K RPM Drive Other Hardware 4 Alteon ACEswitch 180 System Software Operating System Windows2000 Advanced Server Other Software None File System Type NTFS HTTP Software Vendor Microsoft HTTP Software IIS 5.0 and SWC 2.0 # of "threads" 1 Server Cache Microsoft Scalable Web Cache 2.0 Log Mode Binary CLF Network # of Controllers 8 Network Controller Alteon ACEnic PCI # of Nets 8 Type of Nets Jumbo Frame Gigabit Network Speed 1000Mb/s MSL (sec) 30 Time-Wait (sec) 60 Load Generators # of Generators 24 # of Processes Each 16 Model Dell Precision 410 Processor 450MHz PentiumII # of Processors 2 Memory 128MB Network Controller Alteon ACEnic PCI Operating System Microsoft Windows NT 4.0 Workstation SP4 Compiler Microsoft Visual C++ 5.0 Benchmark Configuration Max Requested Load 23100 Max Fileset Size 3319 Test Sponsor Tested By Dell Computer Corp. License Number 55 Test Date Jul-99 =============================================================================== Network Notes <> Each ACEswitch 180 divided into two virtual LANs. <> Each virtual LAN connects one card in the SUT and 3 clients. <> MTU size = 9000 bytes. Other Notes <> IIS settings: listen port 80, no logging, maximum performance. <> IIS Home Directory: Access for Read and Directory Browsing Allowed only, Application Settings permissions "none". <> Network interrupts affinitized one NIC per CPU. <> One disk for OS and pagefile, 4 disks OS striped for web pages, 4 disks PERC striped for logs. <> Registry Settings: <> - Tcpip/Parameters/MaxHashTableSize=0x10000, sets TCB hash table size to 65,536 entries, default is 512. <> - Tcpip/Parameters/TcpTimedWaitDelay=0x3c, sets TIME_WAIT parameter to 60 seconds (non-RFC 1122), default is 240. <> - Tcpip/Parameters/MaxFreeTwTcbs=1400000, sets maximum number of TIME_WAIT tcbs to 1,400,000. <> - Tcpip/Parameters/MaxSendSegments = 0x1000, sets maximum TCP segments sent consecutively to a destination, default 64. <> - Tcpip/Parameters/NumTcbTablePartitions = 32, sets number of independent TCB table segments, default 4. <> - NDIS/Parameters/ProcessorAffinityMask=0, forces processor that handled interrupt to also handle that DPC, default is any processor. <> - Inetinfo/Parameters/ObjectCacheTTL=0, sets IIS object cache timeout to 0 seconds, default 10 minutes <> - Inetinfo/Parameters/OpenFileInCache=0, sets number of open files in object cache to 0 <> - Inetinfo/Parameters/MaxPoolThreads=1, sets pool threads per processor, default is 10 <> - Inetinfo/Parameters/PoolThreadLimit=8, sets total pool threads per system to 8 <> For each Alteon NIC: <> - Altndis(X)/Parameters/SendCoalMax=80, sets number of send buffer descriptors in send ring, default 333. <> - Altndis(X)/Parameters/RecvCoalMax=20, sets number of frames received before notifying host, default 333. <> - Altndis(X)/Parameters/SendCoalTicks=1000, sets number of microsec. before send buffers are returned to host, default 2. <> - Altndis(X)/Parameters/RecvCoalTicks=1000, sets number of microsec. before recv buffers are given to host, default 2. <> - Altndis(X)/Parameters/JumboFrames=1, enables Jumbo Frame support. <> Alternate HTTP/HT.c used. Repeated client-side calls to gethostbyname() omitted, as approved <> by SPEC. No change in measured server performance but reduction in client load. ===============================================================================