16:19:59 #startmeeting opnfv hackfest breakout 16:19:59 Meeting started Tue Mar 15 16:19:59 2016 UTC. The chair is billyoma. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:19:59 Useful Commands: #action #agreed #help #info #idea #link #topic. 16:19:59 The meeting name has been set to 'opnfv_hackfest_breakout' 16:20:14 #info Billy, Al, Tom, Larry 16:21:15 Tom talks about priority based on vlan pcp 16:27:51 Al suggests using multiple interfaces to the guest as a way of prioritizing traffic to the guest 16:28:37 Tom mentions the performance improvemnt to the guest by using vhost-multiqueue. 16:29:41 vhost-multiqueue does not reorder packets on the guest i/f. 16:32:36 Al tcp re-order of adjacent packets is acceptable. It is only when reorder is 3 pkts or more wide that TCP alg. causes issues. 16:43:20 Billy points out the OF pipeline actions/routing are all combined into a single entry in the mini/megaflow OVS lookup. 16:43:44 Billy suggests using NIC multiqueue as the prioritization mechanism. 16:47:35 #question do the NIC queues show up as separate pcie devices? This would require several pmd threads for each queue. 16:49:50 #info Al mentions new OF feature - auxilarry controllers. Want to know about directing priorit packets to auxillary controllers. 16:56:43 #info Al points out that we need to prioritize important traffic both when we know how to deal iwth it (emc...OF hit) and when we do not - packet_out req to controller. 17:41:03 #topic REview of Rel C priority list 17:41:06 #info https://wiki.opnfv.org/release_c_planning/priority_list 17:55:11 #info https://wiki.opnfv.org/release_c_planning/priority_list 18:13:25 #topic Apex Feature List 18:14:14 #info NSH VxLAN based on Pritesh patches 18:14:30 for both ovs-dpdk and kernel dp. 18:22:04 #info Waiting for kernel patches to upstream: Linux VxLAN and LWT (light weight tunnels) 18:31:03 #info pritesh patches only work for kernel 3.12? to 3.17 and need to be rebased to latest kernel 18:32:44 #help 18:32:48 #erase 18:32:51 #delete 18:34:49 #undo 18:34:49 Removing item from minutes: 18:34:56 #undo 18:34:56 Removing item from minutes: 18:36:42 #info pritesh patches to OVS are now abandoned and to be replaced by same functionality in the pending kernel patches. 18:39:44 #topic MPLS over GRE 18:41:20 #info Simon Hormon is working on this. Again there are two sets of patches for this. Kernel and OVS. 18:41:50 #info It's not certain that it will make it in time for Rel -C. 18:42:11 #info Latest info from Simon Horman regarding MPLSoGRE: “At this stage I'm trying to shake out how the vport support for the datapath will look[1] as that will effect both the kernel and user-space implementation of L3 GRE. I would wait for that to be resolved before exercising the code. 18:42:11 18:25 [1] https://www.mail-archive.com/dev@openvswitch.org/msg56967.html 18:42:24 #info https://www.mail-archive.com/dev@openvswitch.org/msg56967.html 18:47:38 #topic 100G, Virtio Container requirement 18:47:55 #info no details on page for these requirements 18:48:06 #topic port placement on cores 18:48:35 #info no details on page for this req - but some work in ongoing for NUMA aware pmd and VM placement 18:48:44 #topic JOID features 18:51:00 #info very little detail or discussion on these items. 18:52:11 #info Some updates to the wiki have been made during this discussion. 18:52:22 #action Billy to forward flipchart photos. 18:52:40 #endmeeting