[ofa-general] Tools for Infiniband Performance measurement
Jeff Squyres
jsquyres at cisco.com
Tue Mar 20 14:44:38 PDT 2007
Note that NetPIPE 3.7 will only use the first port of the first HCA
in your machine (which is a fairly common configuration).
I have submitted a patch to the NetPIPE maintainers that allows you
to specify on the command which port / HCA to use for testing.
On Mar 20, 2007, at 5:43 PM, Kyle Schochenmaier wrote:
> Dukle, Kapil (GE Healthcare) wrote:
>> Hi all,
>> I was wondering if there are any tools specific for Infiniband-
>> performance measurement - throughput, round trip time with adjustable
>> packet sizes.
>>
> The latest NetPIPE benchmarking utility has Infiniband support:
> It will measure what you asked.
>
> http://source.scl.ameslab.gov/NetPIPE/NetPIPE_3.7.tar.gz
>
> Kyle
>> ---------------------------------------------------------------------
>> ---
>> *From:* Dukle, Kapil (GE Healthcare)
>> *Sent:* Monday, March 19, 2007 5:37 PM
>> *To:* general at lists.openfabrics.org; openib-general at openib.org
>> *Subject:* ibping fails in loopback mode
>>
>> Hi all,
>> I'm having trouble getting ibping to work in loopback node. The
>> original config is to connect first port of Blade 1 with first
>> port of Blade 2.
>> For my loopback test, I have connected the Infiniband cable
>> between the 2 ports on the same Linux blade.
>> Both opensm and ibping (server mode) seem to be running on the
>> blade. Ping succeeds to the first active port, but fails
>> for the second one. See below...
>> Why is the ibping to the second active (and linkup) port
>> failing? Am I missing something? The version is OFED1.0
>> *[root at XXXX]# ps -elf | grep opensm*
>> 4 S root 3078 1 0 76 0 - 19039 stext Mar16 ?
>> 00:00:09 */usr/sbin/opensm*
>> 4 S root 5444 5424 0 77 0 - 13981 pipe_w 17:22 pts/1
>> 00:00:00 grep opensm
>> *[root at XXXX]# ibping -v -S &
>> *[1] 5445
>> [root at vre sdc]# ibwarn: [5445] ibping_serv: starting to serve...
>> *[root at XXXX]# ps -elf | grep ibping*
>> 4 S root 5445 5424 0 77 0 - 1454 - 17:22 pts/1
>> 00:00:00 *ibping -v -S*
>> 4 S root 5447 5424 0 77 0 - 13982 - 17:22 pts/1
>> 00:00:00 grep ibping
>> *[root at XXXX]# sminfo*
>> *sminfo: sm lid 0x3* sm guid 0x3ba00010027d9, activity count
>> 203459 priority 1 state SMINFO_MASTER 3
>> *[root at XXXX]# ibstat*
>> CA 'mthca0'
>> CA type: MT25208 (MT23108 compat mode)
>> Number of ports: 2
>> Firmware version: 4.7.400
>> Hardware version: a0
>> Node GUID: 0x0003ba00010027d8
>> System image GUID: 0x0003ba00010027db
>> *Port 1:
>> State: Active
>> Physical state: LinkUp
>> Rate: 10
>> Base lid: 3*
>> LMC: 0
>> SM lid: 3
>> Capability mask: 0x02510a6a
>> Port GUID: 0x0003ba00010027d9
>> *Port 2:
>> State: Active
>> Physical state: LinkUp
>> Rate: 10
>> Base lid: 1*
>> LMC: 0
>> SM lid: 3
>> Capability mask: 0x02510a68
>> Port GUID: 0x0003ba00010027da
>> *[root at XXXX]# ibping -v 3*
>> ibwarn: [5452] ibping: Ping..
>> ibwarn: [5445] ibping_serv: Pong: vre.(none)
>> *Pong from vre.(none) (Lid 0x3): time 0.111 ms*
>> ibwarn: [5452] ibping: Ping..
>> ibwarn: [5445] ibping_serv: Pong: vre.(none)
>> Pong from vre.(none) (Lid 0x3): time 0.087 ms
>> ibwarn: [5452] ibping: Ping..
>> ibwarn: [5445] ibping_serv: Pong: vre.(none)
>> Pong from vre.(none) (Lid 0x3): time 0.069 ms
>> ibwarn: [5452] report: out due signal 2
>> --- vre.(none) (Lid 0x3) ibping statistics ---
>> *3 packets transmitted, 3 received, 0% packet loss, time 2320 ms*
>> rtt min/avg/max = 0.069/0.089/0.111 ms
>> *[root at XXXX]# ibping -v 1*
>> ibwarn: [5449] ibping: Ping..
>> *ibwarn: [5449] main: ibping to Lid 0x1 failed*
>> ibwarn: [5449] ibping: Ping..
>> ibwarn: [5449] main: ibping to Lid 0x1 failed
>> ibwarn: [5449] ibping: Ping..
>> ibwarn: [5449] main: ibping to Lid 0x1 failed
>> ibwarn: [5449] report: out due signal 2
>> --- (Lid 0x1) ibping statistics ---
>> *3 packets transmitted, 0 received, 100% packet loss, time 11358 ms*
>> rtt min/avg/max = 0.000/0.000/0.000 ms
>> Thanks,
>> Kapil
>> !DSPAM:46003995205571657414402!
>> ---------------------------------------------------------------------
>> ---
>>
>> _______________________________________________
>> general mailing list
>> general at lists.openfabrics.org
>> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
>>
>> To unsubscribe, please visit http://openib.org/mailman/listinfo/
>> openib-general
>>
>> !DSPAM:46003995205571657414402!
>>
>
>
> --
> Kyle Schochenmaier
> kschoche at scl.ameslab.gov
> Research Assistant, Dr. Brett Bode
> AmesLab - US Dept.Energy
> Scalable Computing Laboratory
> _______________________________________________
> general mailing list
> general at lists.openfabrics.org
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
>
> To unsubscribe, please visit http://openib.org/mailman/listinfo/
> openib-general
--
Jeff Squyres
Cisco Systems
More information about the general
mailing list