15:59:47 #startmeeting vswitchperf 15:59:47 Meeting started Wed Jan 23 15:59:47 2019 UTC. The chair is srao. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:59:47 Useful Commands: #action #agreed #help #info #idea #link #topic. 15:59:47 The meeting name has been set to 'vswitchperf' 16:00:30 #topic Development Updates. 16:00:52 #info Al Morton 16:02:55 #info Sridhar: Thanks for reviews. 1 more review is pending - Timestamping-Logfiles. 16:03:37 #info Opened a Jira for Reinstalling OS on NOde-1 and Node-2 - and prepare the sandbox. 16:04:28 #info The goal is to create/modify jenkins job to run build process on the new-sandbox. 16:05:08 #info I have started work on "live Results" - spirent and t-rex. 16:05:43 #info Live Results: For every sec (configurable duration) obtain (configurable) metrics. 16:06:39 #topic ONS-NA 2019 16:06:58 #info We have planned a demo session under LFN-booth. 16:07:12 #info https://etherpad.opnfv.org/p/Booth_Demo_Proposals_for_ONS_NA_2019 16:07:58 #info Call for participation: Anybody else planning to attend and participate please let me know. 16:14:06 #info Skydive can be used to display topology and corresponding metrics 16:16:29 #info Felician: Grafana has a dashboard for topology display, we should consider it. 16:16:31 #info Barometr project has used Grafana 16:16:58 #topic Tipsy Update 16:18:42 #info Tipsey - also tests control plane, dynamic control workload effects on dataplane (stress) 16:20:31 #info Output Metrics - Performance Metrics (throughput, latency) 16:20:48 #info always look at smallest packet size, and plot in Mpps (not Mbps) and Latency 16:22:09 #info Dimensions include number of available CPUs, others... 16:22:33 #info Input Metrics - Available number of CPUs, Pipeline Size - spatial and temporal, active-flow count. 16:23:22 #info have pipeline of number of flows in the tables of the switch, another dimension of characterization 16:25:03 #info deployment model fo Tipsey: SW installed on both DUT and SW generator - assumes SW switch is already running! 16:26:37 #info #info have some custom SW to setup the vswitch, etc. and Tipsey on the DUT runs the OpenFlow COntroller 16:26:40 #info Deployment: Tipsy on DUT: Runs on the OFC, , Tipsy on Tester: Orchestrates the measurements, generate traffic needs to be sent. 16:27:28 #info can test HW switch, control with OpenFlow 16:29:00 #info so, the summary is that Tipsey operates the traffic generator (--tgen mode on VSPERF) 16:29:56 #info VSPERF is welcome to re-use Tipsey Piplines, like Mobile Gateway with a SW switch 16:30:34 #info if measuring a SW switch, can use the Openflow Tipey application to configure the switch. 16:31:32 #info OF Controller usually runs on DUT/SUT, but uses TCP connections and can be located anywhere. 16:31:34 #info Connect with felician for reuse of the source. 16:32:13 #info Felician is available for questions when considering integration of Tipsey- xxx 16:33:59 #topic Plugfest results update. 16:34:19 #info Sorry : 1,$ s/Tipsey/Tipsy/g 16:36:02 #info for P2P X-NUMA case, Scenario 2 is not realistic, but kept for consistency with PVP 16:37:15 #info Scenario 3 was worse than 2 -- this was a surprise because the PHD is split across nodes. 16:38:23 #info Finding that it's OK to have PMD on diff node than vSwitch 16:47:50 #info Set the trial-duration reasonably - considering transient impairments 16:49:22 #info link for Long-duration Testing: https://wiki.opnfv.org/display/vsperf/Long+Duration+Testing 16:50:54 #info PROX can help in identifying interrupt processing - and in turn detect process that can be the reason for affecting the performance 16:53:35 #info Waiting for input from Xavier for sec-by-sec prox output to file. 16:55:37 #info Should we put the V-1 of vsperf containers in main VSPERF distribution under docker folder...? 16:58:31 #info Submit a patch - following the review, we can decide. 16:59:15 #endmeeting