Re: Internet2 VoIP SIG Call 9/28 "QoS Not Needed"

[Steve Blair <>]

Garret Yoshimi wrote:

> Hi Ben,
>
> Thanks again for doing the session on the VoIP call.  Wanted to add my 
> two cents to Christian's question at the end of the call, and Mike's 
> comment ... institutional backbones (inside the institution) might 
> very well be QoS enabled as you discussed at the end of the call, 
> however, one of the challenges we face is that there's little chance 
> of realistically supporting real end-to-end QoS when entering 
> individual departmental or workgroup domains that are inside the 
> institution, but that are managed by others.  In our case, we do 
> manage a system-wide interactive video application that operates over 
> a QoS enabled backbone, and, over elements that we control all the way 
> to the edge switch; at the edges of our network, we provision an 
> overlay network in order to maintain end-to-end control.
Hi Ben:

Thanks for the talk. We try to identify a practical delay budget for 
voice and video applications. We then try to consume that budget 
sparingly. We also look for products with low latency implementations of 
codecs. In addition we use vlans for traffic isolation and dot1p marking 
for layer 2 prioritization. This usually  provides a good overall user 
experience.

-Steve

> Best.
> garret
>
> Ben Teitelbaum wrote:
>
> > "Enyeart, Michael J" 
> > <>
> >  writes:
> >
> >  
> >
> > > Ben,
> > >
> > >  
> > >
> > > FYI, the Educause 
> > > Net@Edu
> > >  Integrated Communications Strategies group
> > > recently conducted a survey of fifty (leading) universities, deemed 
> > > most
> > > likely to be deploying convergence solutions.  Forty two universities
> > > responded to the survey. 
> > >  
> > >
> > > Of 42 responding universities, 69% said they planned to implement QoS
> > > techniques as part of a converged services rollout.
> > >     
> >
> >
> > Mike,
> >
> > Interesting.  I presume that most or all of those 69% are referring to
> > some sort of separation of traffic on the campus LAN or rationing of
> > known bottleneck links, which is a perfectly sensible approach in many
> > cases.
> >
> > This is a meaning of "QoS" that I specifically defined as out-of-scope
> > for the talk I gave yesterday, which took "QoS" to mean end-to-end,
> > interdomain, IP service assurances.
> >
> > Please *do* share the report when it is finished.  Thanks!
> >
> > -- ben
> >
> >