Performance of FIND server 29 May 91

Conditions: Loops over XFIND COPY TAPE CMS with SHOW DATE. Normally loaded machines (around 14:00) running VMS and accessing CRNVMC with HTTP/TCP/IP. Loops/min Loops/sec One process on one machine at a time: VXCRNA 27 .45 Hz ONLINE 27 .45 Hz One process on each machine, all at the same time: VXCRNA 25 .4 Hz ONLINE 20 VSOC28 18 VSOC06 16 Total 79 1.3 Hz Thee spawned processes each machine, all the same time (12 in all): VXCRNA 49 .4 Hz ONLINE 21 VSOC28 20 VSOC06 19 Total 109 1.8 Hz The last test apparently brought the VM TCP interface close to running out of buffers and losing telnet connections. This is with a version of the daemon which incorrectly gives precedence to incoming new connections over messages on existing connections, and so the number of connections at any one time may have been larger than necessary. The test demonstrates that the server (if not the IBM) behaves well under overload. The throughput figure is sufficient for our estimates of peak load on