#opendaylight-meeting: tws

Meeting summary

1. intro to ONAP (colindixon, 17:18:18)
   1. Brian Freeman introduces ONAP (which is the merger of OpenECOMP and Open-O) (colindixon, 17:18:41)
   2. open sourcing the core of the network automation platform will help reduce costs for each service provider and provide faster/easier ways to bring VNFs into the ecosystem (colindixon, 17:19:35)
   3. brief overview of the architecture including all the marjor components (colindixon, 17:20:21)
   4. the entire page of stuff is a heat template that will be deployed into 11 VMs and each VM contains docker containers, DCAE "unfolds" into a hadoop cluster with another few VMs (colindixon, 17:21:37)
   5. everything but DCAE comes up in ~15 minutes, in about ~30-45 minutes everything is up, there's then robot tests (colindixon, 17:22:03)
   6. there are robot tests that do health tests for all the VMs as well as ones that onboard some sample VNFs and do end-to-end tests (colindixon, 17:23:31)
   7. building the SDNC from OpenDaylight involves some added features inside, e.g., A&AI (colindixon, 17:27:53)
   8. the APPC is the same, but with more things (colindixon, 17:28:25)
   9. DG (directed graphs) are the key points of how you configure, build and deploy interesting things in ONAP (colindixon, 17:30:31)
   10. shows a REST API call node in the directed graph makes calls, in this case, it's calling to SDN-C to then call into APP-C (colindixon, 17:33:02)
   11. example of the simple Admin Interface running on port 8843 (colindixon, 17:33:48)
   12. svlogic.log has log messages from instantiating/running directed graphs (colindixon, 17:34:36)
   13. record nodes in the DGs are what actually create log entries in the log files (colindixon, 17:36:13)
   14. a list of SDN-C Karaf features (colindixon, 17:36:44)
   15. sliapi is the main module, properties and restapi nodes in the the directed graph, dblib talks to MySQL, etc. (colindixon, 17:37:29)
   16. app-c is similar but a bit different, they have some other REST adaptors (colindixon, 17:39:52)
   17. there are similar, but somewhat more app-focuses Karaf features (colindixon, 17:40:02)
   18. app-c is sdn-c + some more things (colindixon, 17:40:11)
   19. they're built from Beryllium-SR2 OpenDaylight (colindixon, 17:41:18)
   20. things that are not in the open source version of SDN-C, some nodes, e.g., BGP/PECP and PCEP backbone control, those are not in open source now (colindixon, 17:42:42)
   21. colindixon asks and brianF confirms that PNFs and VNFs are intentionally converged and not really a relevant distinction unless it's not possible (colindixon, 17:46:17)
   22. among other things BGP, PCEP and things are harder to find demo-able devices to control (colindixon, 17:47:59)
   23. colindixon asks what's included in SDN-C and APP-C that actually get used, Brian says that NETCONF and BGP and PCEP are included from OpenDaylight, they also have adaptors for chef and other things for APP-C (colindixon, 17:49:04)
   24. in general layer 2-3 is doing stuff with NETCONF and CLI adapators (colindixon, 17:49:17)
   25. LuisGomez asks what VES is, AT&T uses it, it's a telemetry format to get information from the network (colindixon, 17:49:54)
   26. the SNMP collector wasn't included, VES over HTTP was used in place of that in part because the demo VNFs use it (colindixon, 17:50:34)
   27. LuisGomez asks how they use Kafka, Brian says it's used in two ways: one is it's ues in handling TOSCA information and spinning up VMs, It's also used for getting information betwen the SDN-C DCAE and A&AI (colindixon, 17:51:26)
   28. BrianF says DMaaP is based on Kafka, will also have file-based pub-sub where files, e.g., 1 GB, might be published and consumed (colindixon, 17:52:15)
   29. APP-C is the generic VNF-M in ONAP (or at least OpenECOMP) (colindixon, 17:53:54)
   30. this is different from ETSi mano to some extent where there are device-specific VNF-Ms (colindixon, 17:54:58)
   31. colindixon asks when can we see code? BrianF says that if you're a platinum member, you can see it now and have been able to for ~2 weeks. (colindixon, 17:56:04)
   32. phrobb notes that their waiting on things being renamed from OpenECOMP and Open-O to ONAP as well as actually merging code in general to a new architectural diagram. (colindixon, 17:56:50)
   33. BrianF says that there will be 63 gerrit repos for the 11 VMs, "it's a little complex to build" (colindixon, 18:00:23)
   34. takes ~20 GB of disk space once you get you .m2 repo and you can't run everything in a single VM, you really need a hosting environent to run it (colindixon, 18:00:53)
   35. they've been using rackspace (mentioned a bunch of times earlier) (colindixon, 18:01:38)

Action items

1. (none)

People present (lines said)

1. colindixon (38)
2. odl_meetbot (3)