[ewg] OFED 1.2 alpha 1 test Report
Trishal Choudhari
trishal.choudhari at qlogic.com
Wed Feb 28 23:27:13 PST 2007
I am using inbuilt command lines of iperf and netperf.
The other details about the server and HCA are already specified in the test report.
For your convenience I am specifying the below
OS: RHEL4 up4
Kernel: 2.6.9-42ELsmp
Architecture: 64-bit
HCA:Mellanox MT25208
OS: SLES10
Kernel: 2.6.16.21-0.8-smp
Architecture: 64-bit
HCA:Mellanox MT23108
Trishal
________________________________
From: Scott Weitzenkamp (sweitzen) [mailto:sweitzen at cisco.com]
Sent: Wed 2/28/2007 6:15 PM
To: Trishal Choudhari; Openfabrics-ewg at openib.org
Subject: RE: [ewg] OFED 1.2 alpha 1 test Report
Interesting, I see no difference, can you share the command-line you are using the measure this? What server and HCA type are you using?
Scott
________________________________
From: Trishal Choudhari [mailto:trishal.choudhari at qlogic.com]
Sent: Thursday, February 22, 2007 9:34 PM
To: Scott Weitzenkamp (sweitzen); Openfabrics-ewg at openib.org
Subject: RE: [ewg] OFED 1.2 alpha 1 test Report
IPoIB and SDP performance is improved.
1. Significant improvement in seen for small packet sizes.
2.Marginal improvement is seen for default and large packet sizes ( Netperf).
We didn't test MPI.
Trishal
________________________________
From: Scott Weitzenkamp (sweitzen) [mailto:sweitzen at cisco.com]
Sent: Thu 2/22/2007 12:55 AM
To: Trishal Choudhari; Openfabrics-ewg at openib.org
Subject: RE: [ewg] OFED 1.2 alpha 1 test Report
Which perfomance is improved? verbs? IPoIB? SDP? MPI?
Scott
________________________________
From: ewg-bounces at lists.openfabrics.org [mailto:ewg-bounces at lists.openfabrics.org] On Behalf Of Trishal Choudhari
Sent: Wednesday, February 21, 2007 10:52 PM
To: Openfabrics-ewg at openib.org
Subject: [ewg] OFED 1.2 alpha 1 test Report
Qlogic team has completed testing of OFED 1.2 alpha 1 on the following hosts
1.RHEL4 UP4 - 64 bit
2.SLES10 -64 bit
the main observations are
1. Performance on RHEL4 is found improved to a great extent than OFED 1.1
2.Performance on SLES10 is found too low compared to that of RHEL4.
3."ibv_uc_pingpong","ibv_ud_pingpong" tests are failing for RHEL 4 up4 and "ibv_send bandwidth" test is failing for RHEL4 UP4 and SLES10
Trishal
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/ewg/attachments/20070301/dd4c4d89/attachment.html>
More information about the ewg
mailing list