SPECweb96 Result

===============================================================================
                                                        ||
 IBM : RS/6000 S70                                      ||   19264  SPECweb96
 IBM : IBM HTTP Server 1.3.4                            ||
                                                        ||
===============================================================================

   Throughput   Response
     ops/sec      msec
      1950         2.4
      3900         2.6
      5851         2.8
      7801         3.1
      9751         3.4
     11701         3.8
     13761         4.6
     15603         5.0
     17539         6.4
     19264        12.3

===============================================================================

                          General Availability Dates

 Hardware            Aug-98
 Operating System    Aug-99
 HTTP Software       Aug-99

                               System Hardware

 Vendor              IBM
 Model               RS/6000 S70
 Processor           262 MHz RS64-2
 # of Processors     12
 Primary Cache       64KBI+64KBD on chip
 Secondary Cache     8MB(I+D) off chip
 Other Cache         None
 Memory              8192MB
 Disk Controllers    Wide SCSI I/O Controller
 Disk Subsystem      4 9.1GB 16 Bits SCSI Disk Drive
 Other Hardware      ForeRunner LE 155 ATM Switch, IBM 8260 ATM Switch

                               System Software

 Operating System    AIX 4.3.2 + APAR IX86737
 Other Software      None
 File System Type    JFS

                                HTTP Software

 Vendor              IBM
 HTTP Software       IBM HTTP Server 1.3.4
 # of "threads"      23
 Server Cache        FRCA (Integrated)
 Log Mode            Binary Common Log Format

                                   Network

 # of Controllers    23
 Network Controller  23 IBM PCI 155 Mbps ATM Adapters
 # of Nets           23
 Type of Nets        ATM
 Network Speed       155 Mb/sec
 MSL (sec)           30
 Time-Wait (sec)     60

                               Load Generators

 # of Generators     23
 # of Processes Each 12
 Model               23 RS/6000 7043-260
 Processor           200 MHz Pwr-3
 # of Processors     2
 Memory              1Gb
 Network Controller  IBM PCI 155 Mbps ATM
 Operating System    AIX 4.3.2
 Compiler            IBM Cset++ Version 3.1.3

                            Benchmark Configuration

 Max Requested Load  19500
 Max Fileset Size    3051

                                 Test Sponsor

 Tested By           IBM Austin
 License Number      11
 Test Date           Feb-99

===============================================================================

                            Operating System Notes

<> APAR IX86737 - AIX FRCA support

                                Network Notes

<> ATM MTU=9180
<> 1 client per network

                                 Other Notes

<> server no (network option) tuning:
<> no -o sb_max=262144 (max buffer size allowed for a socket, default=65536)
<> no -o somaxconn=8192 (max listen backlog, default=1024)
<> no -o nbc_max_cache=100000 (max file size for network buffer cache, default=1048576)
<> no -o nbc_seg_limit=10000 (number of entry allowed in private segment, default=0)
<> no -o tcp_sendspace=28000 (default socket send buffer size, default=16384)
<> no -o tcp_recvspace=28000 (default socket receive buffer size, default=16384)
<> no -o tcp_timewait=5 (ensure that TIME_WAIT is at least 60 seconds, default=1)
<> Bind FRCA logging thread to cpu 0
<> FRCA Log Format - BINARY
<> processor affinity - bind 23 atm interrupts to 12 cpu (1 intr to cpu-0, 2 intr to cpu 1 - 11)
<> client no (network option) tuning:
<> no -o tcp_sendspace=28000 (default socket send buffer size, default=16384)
<> no -o tcp_recvspace=28000 (default socket receive buffer size, default=16384)
<> no -o delayack=3 (delay ack for connection setup and shutdown, default=0)
<> no -o delayackports={80} (delay ack ports, default={})
<> Alternate HTTP/HT.c used. Repeated client-side calls 'gethostbyname()' ommitted, as approved by SPEC. No change in measured server performance but reduction in client load.

===============================================================================