#opendaylight-meeting: MD-SAL Weekly meeting

Meeting summary

1. 1. moiz - what is the need of habving multiple transaction per thread? (jmedved, 16:13:05)
   2. a thread creates a lot fo transaction w/o need for them to finish (alagalah, 16:13:47)
   
   
2. Example Infinispan Datastore Implementation (edwarnicke, 16:14:05)
   1. a transaction is created per event, and transactions are not tied to a thread (jmedved, 16:14:12)
   2. moiz wants to understand tha if he processes notifications on a songle thread, what is the use case for creating multiple transactions per thread? (jmedved, 16:15:06)
   3. tony responds that a start manager can be updating multiple parts of the subtree in different transactions at different times (jmedved, 16:15:52)
   4. tony: old subtree/ new subtree - nformation about new subtree root is captured in listerner, not transaction (jmedved, 16:22:16)
   5. moiz: when a transaction is created you can not tell which part will be changed, therefore you need to capture the whole tree. but you can re-create the original subtree from the changed subtree (jmedved, 16:23:11)
   6. ewarnicke: it is incredibly useful to have the original subtree. get a sense of what is going on, what changed, etc. (jmedved, 16:24:04)
   7. moiz: data change should be in the event, (jmedved, 16:24:53)
   8. tony: good to have both old and new subtree available in event,s then listener can be stateless; very hard to write binding aware listener that acts only on the delta; for example flow change, you need to know how the flow looked before and after (jmedved, 16:25:56)
   9. moiz no denyoing that data is useful; but wer may have to optimize this, since it can be a lot of data; (jmedved, 16:26:45)
   10. raghu - don’t know which part is modified, so you may have to take a snapshot of the whole subtree (jmedved, 16:27:13)
   11. tony: use instance identifier you can determine the subtree which is being changed (jmedved, 16:27:43)
   12. basheerudin - registrations happen at different levels, so you may have to waklk the whole tree (jmedved, 16:28:07)
   13. tony: for instance identifiers, you can determine exactly what changed and when (jmedved, 16:29:13)
   14. moiz problem is creating a nomralized node from the tree cache. when you create an infinispan transaction you create a normalized node and keep it ready and it has to be from the root (jmedved, 16:30:17)
   15. tony: you can do it from instance id which are listeners and root f the subtree (jmedved, 16:30:50)
   16. tony: data change event you are returning different subtrrees to different listeners based on their registrations (jmedved, 16:31:24)
   17. moiz create multiple of normalized node instances (jmedved, 16:31:55)
   18. Ed Warnicke begins new topic (icbts, 16:37:15)
   19. ewarncike how to facilitate involvement in md-sal issues (jmedved, 16:37:20)
   20. how can we organize work so that chunks can be pulled by different folks (jmedved, 16:38:01)
   
   
3. https://docs.google.com/a/noironetworks.com/document/d/1GdvAiPVMNI1VChojegP-HTrNZznNGvgx2a0I6j7hGdk/edit (alagalah, 16:38:10)
   
4. ewarncike how to facilitate involvement in md-sal issues (alagalah, 16:38:20)
   1. how can we organize work so that chunks can be pulled by different folks (alagalah, 16:38:36)
   2. edwarnicke asks "Who is ready to pick up a shovel?" (alagalah, 16:39:46)
   3. (hard to tell who is responding on WebEx) edwarnicke: Folks from Brocade want to help (alagalah, 16:40:17)
   4. ewarnicke: who has interest working on a datastore? (jmedved, 16:41:20)
   5. arash eghtesadi working on a k-v data store, would like to have a mentor for a better view, ask questions (jmedved, 16:41:57)
   6. raghu: need a series of POCs to validate different assumptions (jmedved, 16:43:25)
   7. I would like to be involved, need to figure out at what level and time commitment. (dbainbri, 16:45:32)
   8. - same like dbainbri - would like to listen in initially & figure out the level and time commitment later (abhijitkumbhare, 16:52:18)
   9. Bug list for MD-SAL: https://bugs.opendaylight.org/buglist.cgi?list_id=7399&short_desc=Milestone&resolution=---&query_format=advanced&short_desc_type=allwordssubstr&component=mdsal&product=controller (edwarnicke, 16:52:48)
   10. (sorry IRC client was crashed) (mlemay, 16:53:02)
   11. https://bugs.opendaylight.org/buglist.cgi?list_id=7399&short_desc=Milestone&resolution=---&query_format=advanced&short_desc_type=allwordssubstr&component=mdsal&product=controller - Current MD-SAL Bugs (edwarnicke, 16:53:58)
   12. . Okay. I noted David and Abhijit as interested in data store. Anyone else I missed? (raghu67, 16:53:58)
   13. md-sal milestone bugs: 688, 629, (jmedved, 16:54:10)
   14. i will be interested in joining datastore and cluster calls (LuisGomez, 17:01:16)
   15. Luis. Okay. Noted (raghu67, 17:01:41)

Action items

1. (none)

People present (lines said)

1. jmedved (37)
2. alagalah (15)
3. edwarnicke (11)
4. odl_meetbot (7)
5. djx (5)
6. abhijitkumbhare (4)
7. phrobb (3)
8. raghu67 (2)
9. icbts (2)
10. dbainbri (1)
11. LuisGomez (1)
12. mlemay (1)