#opnfv-cperf: cperf
Meeting started by dfarrell07 at 19:03:51 UTC
(full logs).
Meeting summary
-
- Lots of ODL perf testing going on this week to
prep whitepaper for ONS (dfarrell07,
19:04:41)
- https://docs.google.com/document/d/10bYHHnMU5pSBJvNON9uVheRq-E1pqGjqp-fIIRZX_yo/edit?usp=sharing
ODL perf whitepaper being created for ONS (dfarrell07,
19:05:20)
- Al LuisGomez and dfarrell07 will be at ONS next
week (dfarrell07,
19:17:16)
- Nikos suggested a "what is sdn perf; which
metrics; which tools" panel for OPNFV summit (dfarrell07,
19:17:46)
- ACTION: dfarrell07
will submit a OPNFV panel CFP for perf topics (dfarrell07,
19:18:10)
- LuisGomez points out that ODL perf is about the
same as other SDN controllers (lessons from this week's whitepaper
efforts) (dfarrell07,
19:18:54)
- dfarrell07 and LuisGomez point out that a big
diff b/t ODL and other controllers is how many features ODL supports
- ODL openflow plugin is about the same as other controllers
(dfarrell07,
19:19:32)
- Al and LuisGomez talk about getting CPerf
feedback to IETF group Al leads (dfarrell07,
19:19:52)
- Marcus is working on perf testing intensely
this week, not on call (dfarrell07,
19:20:09)
- jamoluhrsen has been working on clustering and
a bit of perf stuff for ONS (dfarrell07,
19:20:22)
- ACTION: dfarrell07
needs to get CPerf in OPNFV C release (dfarrell07,
19:21:53)
- https://tools.ietf.org/html/draft-bhuvan-bmwg-of-controller-benchmarking-00
(LuisGomez,
19:22:38)
- Nikos and Al have continued email-based
discussion about perf testing methods/metrics/extensions to
tools (dfarrell07,
19:23:28)
- Nikos has been working on next paper - will
compare Li vs Be perf (dfarrell07,
19:26:52)
- Nikos is seeing much better switch scale data
li->be (dfarrell07,
19:27:23)
- discussion about OVS versions - 2.3 is fine,
but ODL upstream is seeing issues with 2.4 and 2.5 (scale)
(dfarrell07,
19:35:25)
- upstream ODL is working on >2.3 scale
issues, LuisGomez will raise hell about it on OFP call next week to
encourage progress (dfarrell07,
19:36:00)
- discussion about NB tests Nikos is using,
LuisGomez points out that they are likely more realistic/useful than
reactive flow programming from SB (dfarrell07,
19:36:26)
- we may not meet next week due to ONS
(dfarrell07,
19:36:41)
Meeting ended at 19:36:45 UTC
(full logs).
Action items
- dfarrell07 will submit a OPNFV panel CFP for perf topics
- dfarrell07 needs to get CPerf in OPNFV C release
Action items, by person
- dfarrell07
- dfarrell07 will submit a OPNFV panel CFP for perf topics
- dfarrell07 needs to get CPerf in OPNFV C release
People present (lines said)
- dfarrell07 (24)
- collabot` (4)
- LuisGomez (2)
- jamoluhrsen (1)
Generated by MeetBot 0.1.4.