#opnfv-vswitchperf: OPNFV VSPERF
Meeting started by trevor_intel at 14:45:07 UTC
(full logs).
Meeting summary
-
- Trevor Cooper (trevor_intel,
14:45:17)
- Al Morton (acm,
15:01:42)
- 2. Release milestones update (trevor_intel, 15:03:15)
- Need to merge TREX integration and new TCs from
Martin (trevor_intel,
15:05:25)
- patches for Trex integration and Martin's JIRA:
VSPERF-518 should be completed by Aug 11 MS5 (acm,
15:06:51)
- Any documentation updates? (acm,
15:07:14)
- Documentation (trevor_intel, 15:07:21)
- Tuning POD work may add some
documentation (trevor_intel,
15:07:39)
- Ixia license (trevor_intel, 15:08:38)
- Mars used Ixia so it works (trevor_intel,
15:09:03)
- Ixia integration work (trevor_intel, 15:10:27)
- Still waiting for description of the work from
Ixia (trevor_intel,
15:10:46)
- Test WG has discussed long duration tests, need
Pod resources for this, logistics with 5 installers are
complex (acm,
15:12:39)
- what are test cases we would like to include?
Long duration means 2 weeks. (acm,
15:13:12)
- Trevor proposed latency test cases ... need to
tune up what we collect to make them useful (trevor_intel,
15:14:25)
- trevor_intel proposes latency tests for long
duration (acm,
15:14:39)
- Al suggests an alternative is many short
duration tests ... repeated ... makes results duration easier
(trevor_intel,
15:17:37)
- look at min max ave collected from many short
duration latency tests (acm,
15:18:53)
- Sridhar ... what aspect of OVS should we focus
on? (trevor_intel,
15:18:53)
- for example varying number of flows
(trevor_intel,
15:19:32)
- Trevor proposes going with Al's suggestion to
use current latency tests and repeat (trevor_intel,
15:25:25)
- follow-up proposal at Test WG meeting this
week (acm,
15:25:53)
- Test API (trevor_intel, 15:26:02)
- need to discuss naming of a few test
cases (acm,
15:28:33)
- need to fix time display (currently UNIX
format) (acm,
15:28:59)
- time for each test should be replrted
seperately (acm,
15:29:29)
- summary, some small problems and still some
things to do, Mars will supply links (acm,
15:30:27)
- https://wiki.opnfv.org/display/vsperf/VSPerf+testcases+report
(marst,
15:30:54)
- Mars described this report on the Wiki - this
is a list of all test cases VSPERF has (acm,
15:32:37)
- does not include ?? This is the list of tests
included in the API (results collected and stored through the
API??) (acm,
15:35:33)
- ACTION: Trevor
schedule a separate meeting to discuss including Morgan or
Serena (trevor_intel,
15:40:52)
- IMIX (trevor_intel, 15:41:42)
- JIRA- 521 (acm,
15:41:56)
- https://jira.opnfv.org/browse/VSPERF-521
(trevor_intel,
15:43:37)
- Martin and Al discuss 520 and 521 with a view
to merge and assign tasks (trevor_intel,
15:45:02)
- ACTION: Martin K and
acm discuss 520 and 521 with a view to merge and assign tasks
(acm,
15:46:43)
- related to JIRA 523, Noisy Neighbor tests, so
integration of Stress VMs is a sub-task of 523 (acm,
15:49:24)
- Sridhar's VMs at
github.com/opensource-tnbt/stressng-images (acm,
15:51:15)
- Sridhar created thiese VMs after discussion
with Stress-ng author, Colin King (acm,
15:51:54)
- ACTION: Sridhar will
put the VM images into OPNFV artifacts (trevor_intel,
15:53:14)
- need to store the VMs with our other VMs, like
VLOOP, but apparently need special write access, trouble
ticket (acm,
15:53:34)
- 523 is a proposal for suffix to test case
name (trevor_intel,
15:54:07)
- ACE Lee has some MoonGen related questions -
should mail to the VSPERF list (acm,
15:57:54)
- Ace Lee would like to proposes fixes to Moongen
... suggest send suggestions and questions to VSPERF mailing
list (trevor_intel,
15:58:03)
- SR's 524 - live results from traffic
Generators (acm,
15:59:14)
- SR's 525 - want to include *collectd* these
all support the bigger task of *correlation* (acm,
16:00:33)
- CI tuning (trevor_intel, 16:03:34)
- has done much to isolate CPUs, want to try this
profile on the Sandbox, ... (acm,
16:04:28)
- ... but this profile is for Centos, most of Pod
uses Ubuntu 16.04 (acm,
16:05:06)
- uses tuneadm profile, have a new one for Centos
7.3, if recommend this to the community then we shouldget similar
numbers (acm,
16:06:25)
- so, documentation needed recommending a
consistent baseline configuration and these are the perf numbers you
should get... (acm,
16:08:20)
Meeting ended at 16:09:31 UTC
(full logs).
Action items
- Trevor schedule a separate meeting to discuss including Morgan or Serena
- Martin K and acm discuss 520 and 521 with a view to merge and assign tasks
- Sridhar will put the VM images into OPNFV artifacts
Action items, by person
- acm
- Martin K and acm discuss 520 and 521 with a view to merge and assign tasks
- UNASSIGNED
- Trevor schedule a separate meeting to discuss including Morgan or Serena
- Sridhar will put the VM images into OPNFV artifacts
People present (lines said)
- acm (34)
- trevor_intel (24)
- collabot (3)
- marst (1)
Generated by MeetBot 0.1.4.