RE: [AVT] Performance management and new protocols

["Gamze Seckin" <gamze.seckin at vidiator.com>]

I like the idea and I think it could be beneficial only with the
condition that there is a very robust and well-defined explanation of
what is expected. Otherwise misleading performance comments could do
more harm than good.

br
gamze

> -----Original Message-----
> From: Alan Clark [mailto:alan at telchemy.com]
> Sent: Thursday, August 05, 2004 7:58 AM
> To: avt at ietf.org
> Subject: [AVT] Performance management and new protocols
> 
> 
> This email re-iterates a point I made at yesterday's AVT session
related
> to
> performance monitoring and new real time protocols.
> 
> The wide range of new protocols, codecs, applications.. that are
emerging
> will (hopefully) be deployed in either enterprise or service provider
> networks.  In order for these deployments to be successful then EITHER
> 
> (i) the transport network (also servers and other aspects of "the
system")
> needs to be 100% perfect (i.e. the network manager will never have to
> diagnose a problem)
> 
> OR
> 
> (ii) there needs to be a management/ performance monitoring framework
that
> is able to estimate application/ user perceived quality and support
> problem
> resolution.
> 
> Obviously, the normal situation is that the network is quite reliable
but
> not perfect and hence we need (ii) in order to deal with the problems
that
> do arise.
> 
> Typically a new protocol, codec or application is developed with
little or
> no thought given to the feasibility of in-service monitoring and
> management.
> There are some obvious exceptions to this - layer 1-3 protocols often
do
> have accompanying MIBs - however above layer 3 this is much less
common.
> 
> Some of the emerging protocols make management/ monitoring more
difficult
> and in some cases almost impossible.  If we gave some thought to the
issue
> of performance monitoring during the definition of a new protocol/
> application then it would be more likely that we could subsequently
manage
> that protocol/ application when deployed in the field.
> 
> The preferable solution to this problem would be to incorporate a
section
> on
> performance monitoring into each draft - explaining what existing
metrics
> would be suitable for performance estimation or defining suitable
metrics.
> At the AVT meeting there was some doubt expressed that requiring
> additional
> "boilerplate" would be effective however (at least in my opinion)
there is
> a
> real risk of new protocols and applications not being deployed if they
are
> subject to performance degradation and they are "unmanageable".
> 
> The distributed monitoring approach typified by the RFC3611 VoIP
metrics
> represents one way forward - the end system is able to measure
application
> performance directly and economically and provide a lightweight push
model
> to get performance data to collection/ reporting applications.
> Conceptually
> each new protocol/ application could define an associated metrics
block
> for
> a common set of reporting protocols.
> 
> Any suggestions, agreement, disagreement, ideas?
> 
> Regards
> 
> Alan Clark
> Telchemy
> 
> 
> 
> _______________________________________________
> Audio/Video Transport Working Group
> avt at ietf.org
> https://www1.ietf.org/mailman/listinfo/avt

_______________________________________________
Audio/Video Transport Working Group
avt at ietf.org
https://www1.ietf.org/mailman/listinfo/avt