CBench tests against the new OF plugin


Abhijit Kumbhare <abhijitk@...>
 

Moved it to the OF plugin list & also added integration team.

Muthu/Ed - I think if CBench simulates fake 1.0 switches alone - and we are doing this test for 1.0 switches alone there would still be a value in testing it - as the new OF plugin supports 1.0 and 1.3 & largely the 1.0 code path is the same as the 1.3 code path with the new plugin. Of course some version dependent code path will not get tested.


Muthukumaran Kothandaraman---12/02/2013 10:08:57 PM---Hi Abhijit,  We can get started on writing a module which installs drop-rule packet-ins and keep the

From: Muthukumaran Kothandaraman/India/IBM@IBMIN
To: Abhijit Kumbhare/San Jose/IBM@IBMUS,
Cc: Anees A Shaikh/Watson/IBM@IBMUS, Anil Gujele/India/IBM@IBMIN, Anilkumar Vishnoi/India/IBM@IBMIN, "Ed Warnicke (eaw)" <eaw@...>, Gowtham Narasimhaiah/India/IBM@IBMIN, "Prasanna Huddar" <prasanna.huddar@...>, Yugandhar Sarraju/India/IBM@IBMIN
Date: 12/02/2013 10:08 PM
Subject: Re: CBench tests against the new OF plugin




Hi Abhijit,

We can get started on writing a module which installs drop-rule packet-ins and keep them ready. But, as I have seen last, CBench simulates  fake 1.0 switches and NOT 1.3 switches.
So, we could only do a dry-run for 1.0 fake-switches which in my opinion could reveal very basic issues/bottlenecks. Apart from that, the codepath of 1.3 would not be hit to its fullest glory to reveal other key bottlenecks.

Assuming that this stub module is fully MD-SAL based (no AD-SAL routing via compatibility), we require basic workflow of this module being able to receive packet-in and install flow using MD-SAL FRM. If this entire path is working for 1.0 and 1.3
switches, that would be the logical point where we could confidently embark on testing against CBench.

So, is that the idea ?

Regards

Muthukumaran (Muthu)
L  : 080-49141730

P(existence at t) = 1- 1/P(existence at t-1)



Abhijit Kumbhare---12/02/2013 09:29:15 PM---Muthu,  Ed (cc'ed) would like you to have CBench tests run against the new OF plugin. He will be wri

From: Abhijit Kumbhare/San Jose/IBM@IBMUS
To: Muthukumaran Kothandaraman/India/IBM@IBMIN, Yugandhar Sarraju/India/IBM@IBMIN, Anilkumar Vishnoi/India/IBM@IBMIN, Anil Gujele/India/IBM@IBMIN
Cc: Gowtham Narasimhaiah/India/IBM@IBMIN, Anees A Shaikh/Watson/IBM@IBMUS, "Ed Warnicke (eaw)" <eaw@...>, "Prasanna Huddar" <prasanna.huddar@...>
Date: 12/02/2013 09:29 PM
Subject: CBench tests against the new OF plugin



Muthu,

Ed (cc'ed) would like you to have CBench tests run against the new OF plugin. He will be writing a simple service to just write drop flows for all packet-ins. If need be - perhaps you can train Yugandhar & Anil Gujele to do this task tomorrow.

Thanks,
Abhijit