Hi Al, On the data plane benchmarking , there are a lot of parameters that will dictate whether a given tool works best or not for a given use case. From type of benchmark to benchmark profiles to network profile (VLAN, VxLAN...). My personal suggestion is to start simple, and expand to more fancy benchmarks later if needed.
Benchmarking tools are generally a lot more ready than system under test are. If there is a fully working openstack pod in an opnfv lab, it would be interesting to spend the time to benchmark it. By fully working I mean not only control plane but also dataplane with all the required external interfaces ready to get traffic. Any optimization belongs more to the domain of workload optimization (e.g. tweak openstack flavors) more so than to the benchmarking tools per se - at least this is applicable for full stack benchmarking where the tooling should not be allowed to go and tweak the system under test. I am still missing clarity on the actual state of the CNTT RI and how ready it is to take on benchmarking dataplane traffic. I may not be able to attend the release meeting Tuesday morning due to a conflict in my schedule. Thanks Alec From: <[email protected]> on behalf of Al Morton <[email protected]> Date: Monday, August 31, 2020 at 9:31 AM To: opnfv-tech-discuss <[email protected]>, "[email protected]" <[email protected]> Cc: "David McBride ([email protected])" <[email protected]>, "STEELE, SCOT" <[email protected]> Subject: [opnfv-tech-discuss] September 1 Release Meeting: Requirements Work Group (RWG) Hi OPNFV, This week's Release meeting will focus on some very specific requirement areas, and I ask that everyone with expertise please attend! (PTLs, Committers, Contributors). The table summarizing RWG discussion and feedback is here [0]. We need to finish the triage on networking requirements, for example: req.inf.ntw.03 data plane benchmarking (for throughput, latency, etc.) related requirements req.inf.ntw.07 "support network resiliency"... to link failure? interface failure? switch failure? related to req.gen.rsl.01 for OpenStack components What OPNFV tools can we call-upon to help fill these requirements? WHO are the stakeholders? (the resource question). AND the telemetry and monitoring-related requirements, in the Assurance Requirements Table: req.asr.mon.01 (Functional) req.asr.mon.03 (Monitoring) req.asr.mon.04 (Network) Discussion within the Projects (VSPERF, SAMPLEVNF, NFVBENCH, and Barometer) will continue afterward, but we will try to obtain whatever commitments we can for JERMA. Acceleration features (Cyborg) and NUMA-node isolation capability might be possible to test with the performance tools - VSPERF has done it. But these are "SHOULD" level requirements, AND the OS management APIs are still under development. Al [0] https://wiki.opnfv.org/x/wAVbAw
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#24354): https://lists.opnfv.org/g/opnfv-tech-discuss/message/24354 Mute This Topic: https://lists.opnfv.org/mt/76537602/21656 Group Owner: [email protected] Unsubscribe: https://lists.opnfv.org/g/opnfv-tech-discuss/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
