#opnfv-ovsnfv: Weekly Open vSwitch for NFV meeting (18th July 2016)
Meeting started by mdgray_ at 14:02:59 UTC
(full logs).
Meeting summary
- Roll Call (mdgray_, 14:03:08)
- Mark Gray (mdgray_,
14:03:12)
- Tom Herbert (tfherbert1,
14:03:18)
- Chunghan Lee (lch,
14:03:22)
- Billy O'Mahony (billyo,
14:04:37)
- Yardstick Results for OVSNFV with Fuel
(mdgray_,
14:05:19)
- Fuel Update (mdgray_,
14:05:24)
- Apex Update (mdgray_,
14:05:31)
- Spec review (mdgray_,
14:05:51)
- Yardstick Results for OVSNFV with Fuel (mdgray_, 14:06:02)
- Billy has tested Fuel with and without OVSNFV
plugin against yardstick and functest (mdgray_,
14:08:11)
- This is for Rel B (mdgray_,
14:08:20)
- Functest passes (mdgray_,
14:08:27)
- There at least 37 testcases but only 7 were
enabled for CI in Brahmaputra (mdgray_,
14:09:52)
- ACTION: Billy will
follow-up with Yardstick team to understand what tests will be
enabled for Colorado. (mdgray_,
14:10:43)
- TC002 ping test shows 4-times lower latency for
ovsnfv plugin (mdgray_,
14:12:32)
- TC005 File IO test shows broadly similar
results. (mdgray_,
14:14:48)
- TC010 memory latency test shows broadly similar
results. (mdgray_,
14:15:19)
- TC011 Packet delay variation test shows eerily
similar results. (mdgray_,
14:15:56)
- TC012 Mem r/w show better throughput. Unsure
why this would be the case? (mdgray_,
14:19:56)
- TC014 CPU Speed test shows better result for
OVS over OVSNFV. (mdgray_,
14:21:17)
- Many of the tests are run several times and the
results don't change considerably between test runs. (mdgray_,
14:24:38)
- TC037 Pktgen test is showing x10 (2 flows) and
x55 (1M flows) improvement for OVSNFV (mdgray_,
14:25:52)
- This test runs "standard" (i.e. non-dpdk)
pktgen in the guest and sends traffic VM-VM using standard linux
networking interface. (mdgray_,
14:29:20)
- General consensus is that the result look good.
Better performance for the networking tests. Other tests show drops
which may be due to differences in Qemu version and
configuration. (mdgray_,
14:35:45)
- Fuel Update (mdgray_, 14:36:08)
- We are a little behind with our CI (mdgray_,
14:36:25)
- Virtual deployment is work in progress
(mdgray_,
14:37:19)
- Migration to Intel POD has started
today. (mdgray_,
14:37:30)
- Fuel plugin feature has been delivered but not
tested. (mdgray_,
14:38:15)
- No documentation so far. (mdgray_,
14:38:29)
- Apex update (mdgray_, 14:40:15)
- Request from SFC project for ovsnfv to provide
Apex plugin with ovs with NSH (interim patch - not merged
upstream) (mdgray_,
14:42:53)
- RPM will be built by CI and placed in
artifacts (mdgray_,
14:43:13)
- Will not have DPDK support and labelled clearly
(no dpdk and experimental) (mdgray_,
14:43:34)
- When do we merge master branch to experimental
branch? (mdgray_,
14:44:37)
- Suggestion is August 22nd - not sure on
this. (mdgray_,
14:45:03)
- After dates, bug fixes can be cherry-picked
after that point. CI will then swap to the stable branch.
(mdgray_,
14:46:09)
- Fuel plugin will also support NSH
patches. (mdgray_,
14:47:03)
- Clarification, Apex will install OVS with DPDK
aswell (but not with NSH) (mdgray_,
14:48:16)
- Should work with virtual and bare-metal
deployments. (mdgray_,
14:48:34)
- Yardstick should run but no results yet.
(mdgray_,
14:49:10)
- ACTION: Tom: If
results are available for Yardstick, it would be god to compare with
the Fuel results. (mdgray_,
14:49:42)
- Tom asks where are the results stored.
(mdgray_,
14:49:59)
- Billy clarifies that he achieved his results
manually. (mdgray_,
14:50:14)
- Comparable results will give us a good feeling
that things are running correctly. (mdgray_,
14:51:34)
Meeting ended at 14:57:37 UTC
(full logs).
Action items
- Billy will follow-up with Yardstick team to understand what tests will be enabled for Colorado.
- Tom: If results are available for Yardstick, it would be god to compare with the Fuel results.
People present (lines said)
- mdgray_ (48)
- collabot (4)
- billyo (1)
- lch (1)
- tfherbert1 (1)
Generated by MeetBot 0.1.4.