#opnfv-vswitchperf: vSwitchperf
Meeting started by srao at 15:02:54 UTC
(full logs).
Meeting summary
- Development Update. (srao, 15:03:06)
- Integrating VPP-Monitoring with VSPERF.
(srao,
15:04:01)
- Reuse existing project (srao,
15:04:12)
- Monitors VPP and sends metrics over same
Collectd publishing channel. (srao,
15:04:47)
- Use COllectd's 'exec' plugin to grab metrics
from the 'magent' (VPP monitor). (srao5,
15:06:01)
- Requires addition in (a) Grafana (b) Jupyter
Notebook. (srao5,
15:07:13)
- System Configuration APIs are completed - Check
Dependencies, configuration (srao5,
15:08:42)
- Cleanup APIs are completed. (srao5,
15:08:50)
- Al Morton (acmacm16,
15:09:34)
- Start-Test stability improvement
Started. (srao5,
15:09:46)
- Kibana Dashboard for VSPERF Test-Status....
Under exploration. (srao5,
15:13:14)
- ETSI NFV Test WG Chairs is meeting with GSMA in
preparation for F2F meeting. (srao5,
15:18:50)
- Al: Re-look Previous effort - ETSI
INF-010 (srao5,
15:20:38)
- How this RI will be provided? Possibility: A
vendor may offer - A hardware/software vendors.. (srao5,
15:26:18)
- Al: VSPERF should focus on "validating"
Reference implementation - Running tests and confirming with the
"defined" (in the profile) metrics. (srao5,
15:36:24)
- AL: VSPERF should focus on Validating Hardware
reference implementation. Here Hardware really means hardware +
minimal Software. Minimal = OS, Hypervisor, Container-engine,
VSwitch, TestVNF, etc.. (srao5,
15:38:25)
- If SDN Controller inclusion is expected -
VSPERF can work towards that. (srao5,
15:39:44)
- Should VSPERF support in "Creating" Reference
implementation? (srao5,
15:40:58)
- AL: we should be prepared to test it. Creating
may not be its focus. Of course, we can contribute in terms of
Low-Level configuration. (srao5,
15:44:41)
- Question: Once the Ref. Implementation is
validated and Ready - Should VSPERF play role in (a) NVFI
compliance, verification, and cerification of S/W and H/W (b) VNF
compliance validation and cerification. (srao5,
15:48:43)
- GSMA is looking for per VNF-C and vNIC and
vCPU (acmacm16,
15:52:38)
- what range of VNF-C should we test over? we
currently test 1 or 2... (acmacm16,
15:53:32)
- PVP throughput goes down when running
PVVP (acmacm16,
15:54:35)
- PVVP with MemIf does not have this
issue. (srao5,
15:55:07)
- we need the range of VNF scale to test. Per
VNF-C performance will be different if the range is 20 or 50
VNF (acmacm16,
15:55:45)
- Take-away Points: (1) VSPERF play role in
Testing and Validation of a Reference implementation (b) VSPERF can
go beyond baremetal with used with other projects - Functest, YS,
NFVBench. (srao5,
16:03:02)
- VNF Scale Ranges determined by Profile (S,M,L
Tee-shirts), so we have three ranges to test, and someone must
decide the range (acmacm16,
16:04:42)
- CNTT doc
https://cntt-n.github.io/CNTT/doc/ref_model/chapters/chapter03.html#3.3.2
(acmacm16,
16:09:46)
Meeting ended at 16:13:47 UTC
(full logs).
Action items
- (none)
People present (lines said)
- srao5 (20)
- acmacm16 (7)
- srao (5)
- collabot` (3)
Generated by MeetBot 0.1.4.