======================================== #opendaylight-group-policy: ODL-GBP-ARCH ======================================== Meeting started by dconde at 18:06:48 UTC. The full logs are available at http://meetings.opendaylight.org/opendaylight-group-policy/2014/odl_gbp_arch/opendaylight-group-policy-odl_gbp_arch.2014-05-23-18.06.log.html . Meeting summary --------------- * readams says that we don’t need to be so thorough on meeting minutes (tbachman, 18:07:06) * LINK: https://plus.google.com/hangouts/_/calendar/ZHZvcmtpbkBub2lyb25ldHdvcmtzLmNvbQ.bps9id27mhd0pnqqljgjkckoqk?authuser=1 (readams, 18:07:59) * tenant is important under policy (dconde, 18:10:46) * we don't need to virtual ODL instances. (dconde, 18:14:00) * AGREED: we will not subvirtualize JVM or virtualize controller (dconde, 18:15:39) * lenrow wants opp to put a proposal for tenants (dconde, 18:18:18) * do we want another layer of containment? (dconde, 18:20:07) * rather than tenant/ (dconde, 18:20:13) * IDEA: we are talknig of sub-tenants (dconde, 18:21:13) * readsms says at BigSwitch they defined stuff in terms of flow-space. That did not work, although it was a requirement. (dconde, 18:22:50) * it worked fine :-) (readams, 18:23:03) * the answer is blue (tbachman, 18:23:45) * ACTION: lenrow and other will run it by GBP (dconde, 18:24:01) * NB (dconde, 18:24:56) * nothing to be said on that. well discussed a few weeks ago. (dconde, 18:25:23) * scalability constrained by fact of using servlets. (dconde, 18:29:52) * readams says they do a read on a URL and a write to a websockets, hence that's the limitation. (dconde, 18:30:46) * IDEA: XMPP ? is that an option? (dconde, 18:31:05) * no that's how you'd implement notifications over HTTP; this is not implemented in restconf so far as I know (readams, 18:31:12) * uchau question: is there a sequence diagram? (dconde, 18:33:53) * readams says the cloest we have is a picture in front page of architecture. (dconde, 18:34:11) * dvorkini_ says we do need that. (dconde, 18:34:43) * dvorkini_ says different renderers have different semantics. NETCONF, OpenFlow, etc. (dconde, 18:35:34) * uchau asks what determines the renderer? (dconde, 18:35:58) * dvorkinista says rendering down to substrate needs be done in a framework that can accomodate different deliverables. (dconde, 18:36:49) * ovsdb using openflow renderer will do things differently from other methods. (dconde, 18:38:31) * Had to drop due to competing situation. Apologies (lenrow, 18:40:58) * readams a simple renderer can render down to CLI for just using ACLs. (dconde, 18:41:39) Meeting ended at 18:41:44 UTC. Action items, by person ----------------------- * lenrow * lenrow and other will run it by GBP People present (lines said) --------------------------- * dconde (29) * readams (6) * odl_meetbot (5) * tbachman (3) * s3wong (3) * alagalah (2) * lenrow (1) Generated by `MeetBot`_ 0.1.4