====================================== #opendaylight-group-policy: gbp-status ====================================== Meeting started by dconde at 19:59:59 UTC. The full logs are available at http://meetings.opendaylight.org/opendaylight-group-policy/2014/gbp_status/opendaylight-group-policy-gbp_status.2014-05-01-19.59.log.html . Meeting summary --------------- * LINK: for hangout is https://plus.google.com/hangouts/_/calendar/ZHZvcmtpbkBub2lyb25ldHdvcmtzLmNvbQ.nf43n3np1b0kc48it41dsg4v2o (dconde, 20:04:19) * here (mickey_spiegel, 20:06:12) * here (dvorkinista, 20:06:18) * dconde here (dconde, 20:07:21) * here (s3wong, 20:07:24) * start of meeting (dconde, 20:07:52) * agenda item (dconde, 20:07:59) * we see the list of people who will be committed (dconde, 20:08:30) * anyone else should speak up. (dconde, 20:08:37) * AGREED: that the list if what we have for now. (dconde, 20:08:52) * liemmn is 20%, right? but may ramp in future? (dconde, 20:09:27) * liemmn said starting at less than 20 but may ramp to 40% (dconde, 20:09:50) * uchau may ramp up in similar way (dconde, 20:10:03) * Hemanth adds 20% to himself and another TBD from oneConvergence. (dconde, 20:10:34) * A4, a4 in spreadsheet is a reference to tasks in the spreadsheet. (dconde, 20:10:53) * concept of dummy renderer. (dconde, 20:11:11) * we see a diagram on the hangout. (dconde, 20:11:19) * this idagram is going to change since renderers have two components. we have common store for the 4 boxes. repo, ep reg, enforcement repo, op states. (dconde, 20:12:14) * benefits in parallel development by creating a dummy renderer that statically responds (like a stub?) same for SB renderer (dconde, 20:12:45) * it's a stub with a statically defined field. (dconde, 20:13:12) * if I know the subject,etc. we can stuff the subject into the stub (dconde, 20:13:34) * we need to work with the scorer for the scalar generation algorithm. Someone in APIC team has done that. So alagalah will contact them for background. (dconde, 20:14:14) * same for thresholding and acknowledgement (dconde, 20:14:28) * the boxes have very little intelligence in them. (dconde, 20:14:39) * the renderer has all the intelligence in them. (dconde, 20:14:49) * parallel implementation (dconde, 20:15:15) * the DSE talk to the Endpoint registry / mapper (dconde, 20:17:29) * DSE / Governance / Congress /etc. needs to fit in. (dconde, 20:17:45) * DSE stands for data services engine (dconde, 20:18:00) * Plexxi guys contributed that to the Congress project. (dconde, 20:18:10) * DSE pulls out network metadata from other sources (dconde, 20:18:36) * there is a placeholer in the release plan. (dconde, 20:19:36) * we start with OpenFlow and OpFlex moved into incubation this morning (dconde, 20:19:51) * so that's a real project. (dconde, 20:20:03) * we want to implement REnderer common and migrate OF Strawman (dconde, 20:20:58) * we will talk about that in the ARCH meeting tomorrow. (dconde, 20:22:11) * we think subscription and rendering is intertwined. (dconde, 20:22:27) * everyting in column A should be a sheet. (dconde, 20:22:38) * interdependencies are critical. (dconde, 20:22:53) * release plan schedule (dconde, 20:23:13) * we are looking at gates. (dconde, 20:23:24) * M1 to M3 is easy (dconde, 20:23:30) * M4 and 5 are harder (dconde, 20:23:34) * all things are gated on beginning of a week except we have a conflict with Labor day for code freeze (dconde, 20:24:41) * probability for working on labor day is low (dconde, 20:25:24) * dvorkinista reviewing release plan. (dconde, 20:25:36) * this is aggressive. (dconde, 20:25:49) * if people at 20% then if you are not familiar with ODL it's hard. (dconde, 20:26:08) * we can do stuff like pair work, etc to ramp it up. alagalah says we need to experience first hand if this is sane. (dconde, 20:26:50) * if the goal is to render everything forever, it's hopeless. (dconde, 20:27:23) * we need to understand the subset. (dconde, 20:27:30) * dlenrow is talking about stretch goals, etc. a few subjects to render (dconde, 20:27:58) * dvorkinista says if we do it well, it's better than something fast that does not work. (dconde, 20:28:26) * mickey_spiegal says that contracts and such need to be put in early. (dconde, 20:29:07) * dvorkinista says it needs to be fast too. (dconde, 20:29:27) * alagalah says a subset is doable. (dconde, 20:29:48) * dvorkinista (dconde, 20:30:18) * most affects are on service chaining, but dvorkinista says we can hold off on that but it needs to be done well, and we can wait for that. (dconde, 20:30:43) * we need to be very careful and not over commit. (dconde, 20:31:12) * service chaining (dconde, 20:31:25) * looking at diagram (dconde, 20:31:50) * the diagram will be mailed out. (dconde, 20:32:06) * dvorkinista we need to make sure it is extensible. (dconde, 20:32:15) * OK to skip functions to edit without rewrite. (dconde, 20:32:23) * dvorkiinsta not worried about policy. EPR needs to be fast. Renderer is where all crazy behaviour is. (dconde, 20:32:57) * it seems to make sense for parallel development. (dconde, 20:33:18) * alagalah says we need to worry about extensibility in renderer. (dconde, 20:33:37) * we need to describe a policy that governs that. (dconde, 20:33:58) * all of these things can be subclassed, says dvorkinista (dconde, 20:35:10) * OF stuff is down below, says mickey_spiegel (dconde, 20:35:20) * this is in the OF render, not part of the common renderer. (dconde, 20:36:15) * dvorkiista says it's another layer of abstraction. OpenFlow based, affinity, etc. (dconde, 20:37:00) * renderer (dconde, 20:37:34) * we are redrawing diagram. (dconde, 20:37:47) * adding OF renderer below Flow based renderer (dconde, 20:38:01) * idea to make the dummy object is statically defined and we invoke this into flows. (dconde, 20:38:54) * flow based rendering is not stateless. (dconde, 20:39:22) * render common is constantly rendering. Are we worried about passing vs. storing. (dconde, 20:39:58) * we talked about schedule. (dconde, 20:41:29) * recap (dconde, 20:41:35) * is this a sane plan? (dconde, 20:41:42) * keep everything equal. and statically define it. (dconde, 20:43:16) * readams asks.does the dummy version need to be created first? (dconde, 20:43:47) * readams says we cannot make progress until model is defined. (dconde, 20:44:26) * readams what is A5 for renderer? (dconde, 20:45:22) * alagalah asks which parts do you want to contrib to? (dconde, 20:46:09) * call fo volunteers (dconde, 20:46:21) * call for volunteers (dconde, 20:46:31) * we have a proposal now. (dconde, 20:47:34) * liemnm how do we do tests? (dconde, 20:47:52) * we do have dependencies -- says lenrow (dconde, 20:48:40) * alagalah UI extensions are unknown. (dconde, 20:48:50) * we need to look at QOS, openflow 1.3 metering impl, etc. (dconde, 20:50:05) * this may limit what labels we can use. (dconde, 20:50:11) * ACTION: alagalah to follow on edwarnicke ODL plans (dconde, 20:50:31) * test dev platform == mininet. but ideally we need real set-up. (dconde, 20:52:51) * tbachman to document the ovsdb / openstack / ovs dev environment (dconde, 20:53:20) * also for openflow on mininet. (dconde, 20:53:28) * for OpFlex, it's nascent, so we need to figure out. probably mininet too,b ut it may be ovs switches directly on Linux. Who knows. (dconde, 20:53:58) * spredsheet was emailed out. (dconde, 20:54:44) * sharing is failing. (dconde, 20:55:10) * sharing of gDocs done. (dconde, 20:56:13) * anything else? (dconde, 20:56:53) * monday COB has the final commitment. (dconde, 20:57:22) * if you cannot get on ghangout then go to IRC and say you cannot get on, and then we can fire off a WebEx. (dconde, 20:59:07) * of use YouTube live to stream video. (dconde, 20:59:40) Meeting ended at 20:59:53 UTC. Action items, by person ----------------------- * alagalah * alagalah to follow on edwarnicke ODL plans People present (lines said) --------------------------- * dconde (122) * dlenrow (4) * dvorkinista (4) * odl_meetbot (4) * jmedved (2) * tbachman (2) * mickey_spiegel (1) * s3wong (1) * alagalah (0) Generated by `MeetBot`_ 0.1.4