#opnfv-testperf: OPNFV Traffic Profiles

Meeting summary

1. 1. Morgan Richomme (morgan_orange, 13:17:45)
   2. Bryan Sullivan (bryan_att, 13:17:51)
   3. So I uploaded the mobility profile to gerrit, as far as it has been developed (bryan_att, 13:19:02)
   4. There was a request in Yardstick to add a control plane component which is not complete yet, so there is a placeholder (bryan_att, 13:19:47)
   5. https://gerrit.opnfv.org/gerrit/#/c/1312/1/commons/mobile/readme.rst (morgan_orange, 13:21:54)
   6. I'll add some more background description to address the questions (bryan_att, 13:24:22)
   7. The profile is "per average user session" (bryan_att, 13:24:57)
   8. I don't have busy-hour/curve info but I'm not sure how that would help in how we intend to use the profile (bryan_att, 13:25:51)
   9. Tweaking a load generator to increase the load based upon this average, or to simulate load variation over time is something I would say is more test design, not profile (bryan_att, 13:27:31)
   10. The point of this profile is to provide an average traffic mix for an average session (bryan_att, 13:28:11)
   11. For the json description are there any standard formats for that? (bryan_att, 13:29:47)
   12. I was thinking it might help drive more consistent traffic generator behavior if we did have some semantic expression of the profile, but it depends what they support, and it seems there is no standard (bryan_att, 13:31:14)
   13. But I can propose a json representation if one says they can use it (bryan_att, 13:31:55)
   14. http://nfvwiki.etsi.org/index.php?title=HT02_-_Test_Methodology_for_NFV (morgan_orange, 13:34:53)
   15. http://docbox.etsi.org/ISG/NFV/Open/Drafts/TST001_-_Pre-deployment_Validation/ (morgan_orange, 13:35:12)
   16. One more thing; the wiki https://wiki.opnfv.org/opnfv_functional_testing (bryan_att, 13:35:45)
   17. I added a link to the traffic profile page https://wiki.opnfv.org/traffic_profile_management (bryan_att, 13:36:22)
   18. How much do we want to maintain on the wiki vs a document in git? (bryan_att, 13:36:43)
   19. I'd rather just point the wiki to the doc in git, and migrate the wiki content there (bryan_att, 13:37:50)
   20. Yes, I don't want to confuse anyone. I think much of the rationale info as to why there are multiple profiles etc should be in a doc (bryan_att, 13:39:09)
   21. The doc will be auto-generated and linked to the wiki, not scraped (bryan_att, 13:39:32)
   22. It's easy ala the link to the "Current working draft " on Copper (bryan_att, 13:40:20)
   23. Linked from the artifacts page (bryan_att, 13:40:36)
   24. Last thing; I will send out a note to collect the other profiles into the same repo (bryan_att, 13:42:11)
   25. There is an open section in the profile for a control plane profile part (bryan_att, 13:44:49)
   26. For now through the "session management" overhead will be developed as part of turning the profile into a specific test scenarion for a "DUT/SUT" (bryan_att, 13:45:43)
   27. For example, a SGW to Gi-Lan elements to FW/NAT (bryan_att, 13:46:22)
   28. We have to craft a session profile in order to generate the traffic - otherwise we are just blasting data at a certain mix (bryan_att, 13:47:18)
   29. So there is a placeholder, and I will work with Ixia and Spirent on how to describe the session (control plane) profiles (bryan_att, 13:48:18)
   30. If anyone has something they want to bring to the table then great (bryan_att, 13:48:37)
   31. Otherwise it will be like this: A session for a particular application lasts an average time (bryan_att, 13:49:10)
   32. During that session there may be a mix of applications used in varying percentages (bryan_att, 13:49:29)
   33. The applications have certain control plane protocol implications (number/type of messages) per the standards (bryan_att, 13:50:08)
   34. It would be great to get end-user input into the session / control plane profiles (bryan_att, 13:50:55)
   35. I will send a note for input, and also basic design ideas (bryan_att, 13:51:11)
   36. I am not an expert in this but I developed profiles for WAP testing back in the day... (bryan_att, 13:51:38)
   37. And most apps including streaming were included in the tests (bryan_att, 13:52:14)
   38. But mobility traffic is at the same time more complex today (more apps) and more skewed toward one type of app (video) (bryan_att, 13:52:44)
   39. We are probably good with one global profile (bryan_att, 13:53:27)

Action items

1. (none)

People present (lines said)

1. morgan_orange (56)
2. bryan_att (48)
3. collabot (4)
4. Malla_ (1)