Re: REG: draft-vsperf-bmwg-vswitch-opnfv-02

Hi Sudhin,

VSPERF has address caching capacity and address learning rate
(see https://wiki.opnfv.org/display/vsperf/LTD+Test+Spec+Overview)
but this doesn't include clear and re-learning.

You've also listed a series of tests where the performance
measurement is "recovery time", and implicitly determine if
recovery from link flap, NIC reset, reboot, etc., is successful.
But the recovery times may not be very interesting for virtual functions
(so *many* operations have to happen in the case of reboot)
and they may involve manual operations in some cases.

Also, we don't include pass/fail functional tests in BMWG's charter,
we're all about performance.

Al



From: Sudhin Jacob [mailto:[email protected]]
Sent: Wednesday, July 13, 2016 8:49 AM
To: MORTON, ALFRED C (AL)
Cc: Tahhan, Maryam; [email protected]
Subject: RE: REG: draft-vsperf-bmwg-vswitch-opnfv-02

Hi Al,

Apologies for the confusion.

In performance testing, negative scenarios are considered as the paramount . Let me explain, this scenarios includes wide range of tests like link flap, reboot, reset line cards,  scale to maximum mac limits and then forcibly clear the entries and then relearn etc.
In Vswitch which is an instance created in a server so the scenarios can be tailored made to cater the virtual environment. Correct me if I am wrong.

Regards,
Sudhin

From: MORTON, ALFRED C (AL) [mailto:[email protected]]
Sent: Wednesday, July 13, 2016 5:31 PM
To: Tahhan, Maryam <[email protected]<mailto:[email protected]>>; Sudhin Jacob <[email protected]<mailto:[email protected]>>; [email protected]<mailto:[email protected]>
Subject: RE: REG: draft-vsperf-bmwg-vswitch-opnfv-02

Hi Sudhin and Maryam,

Thanks for your review and comments so far.

In addition to Maryam's request below,
I want to ask for an additional clarification, Sudhin,

You last question mentions "negative scenarios".
What sort of things are you imagining?
(the term is unfamiliar to me).

thanks,
Al

From: bmwg [mailto:[email protected]] On Behalf Of Tahhan, Maryam
Sent: Wednesday, July 13, 2016 6:54 AM
To: Sudhin Jacob; [email protected]<mailto:[email protected]>
Subject: Re: [bmwg] REG: draft-vsperf-bmwg-vswitch-opnfv-02

Hi Sudhin

Comments inline below.

Best Regards,
Maryam

From: bmwg [mailto:[email protected]] On Behalf Of Sudhin Jacob
Sent: Saturday, July 9, 2016 6:40 AM
To: [email protected]<mailto:[email protected]>
Subject: [bmwg] REG: draft-vsperf-bmwg-vswitch-opnfv-02

Respected Authors,

The draft is comprehensive and covers most of the aspects the benchmarking parameters. However I have some queries kindly clarify.

In section 5.9 contains matrix of scale  in activation and operation.

Could you please let me know whether benchmarking the  Vswitch performance by  measuring the time taken to learn the maximum capacity of switch mac address, clear it and relearn.

<MT> I'm not sure I understand the question here was there a missing part to the question above? can you please clarify what you are asking?

Could you please let me know the stressful scenarios mentioned in section 2 of the draft.
<MT> Scenarios where you stress the CPU, Memory, scaling the number of flows and VNFs attached to the vswitch, bringing up a number of VNFs together... while the vswitch is in operation.

Could you please let me know, correct me if I am wrong  the Vswitch must be benchmarked with negative scenarios.

Regards,
Sudhin

_______________________________________________
bmwg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/bmwg