#opendaylight-meeting: MD-SAL interest call

Meeting started by tbachman at 16:05:42 UTC (full logs).

Meeting summary

  1. agenda (tbachman, 16:06:16)
    1. Agenda items: tatus update, Recap on Modeling Network Topologies, Bug reporting/Fixing (tbachman, 16:06:52)
    2. colindixon asks to add wildcard queries as opposed to wildcard subscriptions on the MD-SAL (tbachman, 16:07:23)

  2. recap on modeling network topologies (tbachman, 16:09:57)
    1. the discussion offline didn’t happen, here so starting here (allocating 15 minutes) (colindixon, 16:10:58)
    2. the network topology model supports multiple topologies, today we have one openflow-oriented topology with the key “flow:1” (colindixon, 16:11:41)
    3. Ramkumar is sharing screen as he walks through examples (devinavery, 16:12:53)
    4. ramkumar says his understanding is that each link has a reference back to the node (tbachman, 16:14:07)
    5. colindixon says that in the long run we probably won’t have this separation between inventory and topology (tbachman, 16:15:07)
    6. but will be augmentations instead (tbachman, 16:15:15)
    7. ramkumar says this might be better, b/c regardless of the node type (flow-capable, etc.), the topology doesn’t care (tbachman, 16:16:01)
    8. ramkumar asks if there are multiple topologies, and if there are multiple toplogies, how do relate between different nodes (tbachman, 16:17:12)
    9. jmedved says that’s difficult to model, and you need something to reconcile, which would be another plugin (tbachman, 16:17:32)
    10. ramkumar says that whatever’s using different node types has to know how to talk to each of them (tbachman, 16:20:12)
    11. ramkumar says that the plugins (e.g. openflow, BGP, etc.) only need to know how to talk to their nodes (tbachman, 16:20:44)
    12. Jan states that each plugin should create its own topology and then a second plugin should be used to relate two topologies together (devinavery, 16:25:40)
    13. colindixon says you can create an overlay topology which the different toplogies can attach to (tbachman, 16:28:50)
    14. jmedved notes there is a producer and a consumer of the topology, and that the consumers don’t need to understand the details of the producer (tbachman, 16:30:02)
    15. colindixon says you could work on the global topology, which is created by flattening everything (tbachman, 16:30:37)
    16. ramkumar says that if you don’t have a flat topology, how do you know how to navigate between the topologies? (tbachman, 16:30:53)
    17. colindixon says that topologies can have references to other topologies (tbachman, 16:32:28)
    18. and the application can traverse a metatopology that links the two (tbachman, 16:32:39)
    19. colindixon notes that we don’t have interconnections today, but the model supports it (tbachman, 16:33:09)

  3. bugs (tbachman, 16:36:59)
    1. https://wiki.opendaylight.org/view/OpenDaylight_Bugs <= link that devinavery created to help with bugs (tbachman, 16:38:40)
    2. The wiki page describes how to use fields like Target Milestones (tbachman, 16:39:24)
    3. The page has quick links to find bugs for each project in both Helium and (upcoming) Lithium releases (tbachman, 16:40:03)
    4. The bugs listed in the Helium links are the bugs that are intended to be fixed in the Helium release (tbachman, 16:40:43)

  4. MD-SAL and wildcards (tbachman, 16:44:56)
    1. colindixon is trying to get the node in the topology that corresponds to a host with a given IP address (tbachman, 16:49:45)
    2. his first attempt was to get an instance identifier from NetworkTopology’s children (tbachman, 16:50:15)
    3. and that this would provide a list (tbachman, 16:50:22)
    4. however, that read returned a single host node, and not a list (tbachman, 16:50:39)
    5. The second attempt had to iterate through a higher level list (tbachman, 16:51:15)
    6. colindixon wants to know if there’s a way to wildcard the instance identifier (tbachman, 16:51:43)
    7. rovarga says that the datastore has to do precisely the same thing that’s being done in colindixon’s second attempt (tbachman, 16:52:14)
    8. rovarga says that instead of a wildcard, you’d rather have a query language to express what it is you’re looking for (tbachman, 16:53:09)
    9. an advantage of an xpath query is that you can do things like “give me all of the things that have host nodes”, which wouldn’t be available in the query language (tbachman, 16:55:26)
    10. jmedved asks if colindixon could discuss his use case with sharon (tbachman, 16:55:48)
    11. colindixon says that you can’t really do that with SQL because it breaks schemas (tbachman, 16:56:03)
    12. devinavery says that could be a lithium focus: making searching of MD-SAL easier (tbachman, 16:56:36)


Meeting ended at 16:58:03 UTC (full logs).

Action items

  1. (none)


People present (lines said)

  1. tbachman (49)
  2. colindixon (6)
  3. odl_meetbot (5)
  4. devinavery (4)


Generated by MeetBot 0.1.4.