Hi, I have seen large variances between runs since I started SDP work. I have not seen more than 25% (or so) difference with netperf runs, but even that is unacceptable. The change from 3200 to 1200 is big enough that I expect something else is going on. Is it possible that the second run did not use SDP? I always export LD_PRELOAD (or unset LD_PRELOAD) prior to running the tests.
With the publication of the zero copy bcopy changes, bugs and performance issues have moved to the top of my list. I would love to get any specific performance test parameters that seem to reliably cause especially egregious results like these. Jim Mott -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Moshe Kazir Sent: Sunday, October 28, 2007 10:11 AM To: [EMAIL PROTECTED]; [email protected] Cc: Moni Levy; Alon Verner Subject: [ewg] Running netperf and iperf over SDP While running netperf and iperf over SDP I get unstable Performance results. On iperf I get more then 25 % difference between minimum and maximum. On netperf I get the following amazing result. -> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ # while LD_PRELOAD=/usr/lib64/libsdp.so ./netperf -H 192.168.7.172 -- -m 512 -M 1047 ; do echo . ; done TCP STREAM TEST to 192.168.7.172 Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 126976 126976 512 10.00 3247.39 . TCP STREAM TEST to 192.168.7.172 Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 126976 126976 512 10.00 1222.48 . ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Can some one spar me a hint ? What I'm doing wrong ? The test run on x86_64 , sles 10 sp1 , OFED-1.2.5 Moshe ____________________________________________________________ Moshe Katzir | +972-9971-8639 (o) | +972-52-860-6042 (m) Voltaire - The Grid Backbone www.voltaire.com _______________________________________________ ewg mailing list [EMAIL PROTECTED] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/ewg _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
