========================================== #opendaylight-group-policy: DataStore 4/28 ========================================== Meeting started by regXboi at 14:01:44 UTC. The full logs are available at http://meetings.opendaylight.org/opendaylight-group-policy/2014/datastore_4_28/opendaylight-group-policy-datastore_4_28.2014-04-28-14.01.log.html . Meeting summary --------------- * LINK: https://wiki.opendaylight.org/view/Group_Policy:Sub-Groups:DATASTORE (alagalah, 14:02:24) * agenda bashing (regXboi, 14:04:01) * who is contributing and what time can be devoted? (regXboi, 14:04:50) * alagalah sent this to the mailing list (regXboi, 14:05:10) * only regXboi and s3wong have replied so far (regXboi, 14:05:22) * discussion of where performance considerations should be tracked - here or in MD-SAL (regXboi, 14:07:52) * ACTION: regXboi to update wiki page with the scope of datastore (regXboi, 14:10:53) * alagalah: Charter of this group is to determine requirements of GBP datastore and it's mapping to MD-SAL and any changes we may need to MD-SAL ala APIs. This meeting will roll into Arch and DATASTORE will actually, once we get to code, we will probably have only 3 groups (EP-mapper/policy, OperationalState, Renderers) (alagalah, 14:10:59) * ACTION: (not direct for this) jmedved to start actual model discussion (regXboi, 14:15:27) * ACTION: jmedved to produce a YANG model from GBP UML 0.97 with Tony (Pantheon?) and email to the list in next two days. (alagalah, 14:15:39) * currently jmedved uses cbench to hit the datastore currently (regXboi, 14:15:43) * LINK: http://www.joedog.org/siege-home/ (regXboi, 14:16:38) * LINK: https://plus.google.com/hangouts/_/calendar/ZHZvcmtpbkBub2lyb25ldHdvcmtzLmNvbQ.e7ngfp46nqhv5oltibelval6o8 (alagalah, 14:17:46) * Rob Adams asks if we are putting the "cart before the horse" talking about MD-SAL performance metrics (alagalah, 14:18:55) * Before we have an actual implementation to measure (alagalah, 14:19:03) * regXboi claims what we are doing is setting the methodology so it is common when comparing implementations (regXboi, 14:19:25) * regXboi points out that he wants to settle on a methodology so we can use the same methodology to benchmark existing, and then measure our implementations against it and identify improvements etc (alagalah, 14:19:40) * LINK: https://wiki.opendaylight.org/view/OpenDaylight_Controller:MD-SAL:Architecture:DOM_DataStore (alagalah, 14:20:47) * according to jmedved that is the latest documented, but it still needs to be updated with latest infromation (regXboi, 14:21:20) * new in-memory data store is going to an external customer so we'll have field trial data to work with (regXboi, 14:22:27) * jmedved is interest in comparing performance of current tree with mlemay's KV implementation of the in-memory datastore (regXboi, 14:23:13) * (edit) s/interest/interested/ in last info (regXboi, 14:23:34) * alagalah want to clarify the separation of md-sal and in-memory datastore and is asking if readams is comfortable with the implementation implications of that (regXboi, 14:24:28) * readams is still working it (regXboi, 14:24:38) * a bit of a discussion of APIC vs ODL sharding (regXboi, 14:34:11) * alagalah asks jmedved about a) does sharding fall under GBP or MD-SAL. jmedved says MD-SAL. b) Does MD-SAL have a Release Plan that I can sync the GBP release plan with so that when a portion of GBP needs to leverage sharding (and the leader + 2x follower concept from APIC for a shard) that MD-SAL has a prototype ready to go. jmedved points out that MD-SAL is not an official project but a sub-pa (alagalah, 14:42:08) * ACTION: alagalah once he publishes the first cut of the GBP Helium Release Plan will also have a document that gives context to the relationships between subgroups, and other projects, to justify why certain steps are at certain gates in the critical path (alagalah, 14:42:55) * alagalah asks jmedved about a) does sharding fall under GBP or MD-SAL. jmedved says MD-SAL. b) Does MD-SAL have a Release Plan that I can sync the GBP release plan with so that when a portion of GBP needs to leverage sharding (and the leader + 2x follower concept from APIC for a shard) that MD-SAL has a prototype ready to go. jmedved points out that MD-SAL is not an official project but a sub-pa (alagalah, 14:43:59) * rt of the Controller project, and these discussions are still underway. (regXboi, 14:44:28) * (that finished the previous info) (regXboi, 14:44:34) Meeting ended at 14:49:41 UTC. Action items, by person ----------------------- * alagalah * alagalah once he publishes the first cut of the GBP Helium Release Plan will also have a document that gives context to the relationships between subgroups, and other projects, to justify why certain steps are at certain gates in the critical path * regXboi * regXboi to update wiki page with the scope of datastore * **UNASSIGNED** * (not direct for this) jmedved to start actual model discussion * jmedved to produce a YANG model from GBP UML 0.97 with Tony (Pantheon?) and email to the list in next two days. People present (lines said) --------------------------- * regXboi (25) * alagalah (24) * odl_meetbot (4) * s3wong (4) Generated by `MeetBot`_ 0.1.4