Hi Kevin,
In steady state, there will be a scale impact of enabling stability 
measurement. However, it can be fairly small if the implementation is 
efficient. In our prototype, we saw less than 3% impact on CPU performance (and 
hence, little-to-no impact on scale) when we enabled this measurement (we 
monitored both, timestamps and sequence numbers). 
Best,Ashesh


> From: [email protected]
> To: [email protected]; [email protected]
> Subject: RE: Questions on draft-ashesh-bfd-stability-00.txt
> Date: Mon, 30 Mar 2015 13:38:15 +0000
> 
> Hi Santosh,
> 
> Thanks for your responses. Will look the updated version.
> On my last question, yes it is about scale impact in steady state.
> 
> Kevin
> 
> -----Original Message-----
> From: Santosh P K [mailto:[email protected]] 
> Sent: Saturday, March 28, 2015 10:45 AM
> To: D'SOUZA, KEVIN L; '[email protected]'
> Subject: RE: Questions on draft-ashesh-bfd-stability-00.txt
> 
> Hello Kevin,
>      Thanks for your comments. Please see my inline reply. 
> 
> >1) Could you clarify if interop will be supported between these features and 
> >a BFD instantiation that does not support these features
> 
> Yes, it should have a backward compatibility. We will update document on 
> backward compatibility section with use cases. 
> 
> > 2) Can you specify that these features will be optional for an 
> > implementation and some method should be provided to turn this capability 
> > off if needed
> 
> This is optional and it might be turned off by configuration? I think this is 
> implementation detail but we could have recommendation in draft. 
> 
> >3) Can you clarify in the scope the relationship between the measurements 
> >that one would obtain with this method vs other traditional methods to 
> >measure >such criteria (e.g., CFM, TWAMP).  My concern is if this feature 
> >produces measurements that are different from the traditional methods, then 
> >it would be >difficult to reconcile them in a production environment
> 
> BFD will not replace CFM or TWAMP and it is not equipped to do so. We are 
> updating the document with use cases and other recommendation please stayed 
> tuned. 
> 
> > 4) How would this impact the scale of the BFD protocol?
> 
> Do you mean scale impact in steady state? 
> 
> 
> Thanks
> Santosh P K 
> 
                                          

Reply via email to