VOIP and QOS across local network and the Internet

Jo Rhett jrhett at svcolo.com
Tue May 15 23:43:17 UTC 2007


I've never done it on a core switch, but at the edge doing traffic  
prioritization is dead easy.

I imagine that doing it in the core just means that you need to make  
sure your core switch is up to it.  For example, Force10 units are  
likely to handle it, Extreme units are guaranteed to fail, et al.  
what you already know about handling load ;-)

On May 15, 2007, at 11:41 AM, Rick Kunkel wrote:
> We're getting an increasing amount of pressure from VOIP providers
> colocated with us and from VOIP end-users to prioritize traffic on our
> network.  From a network administrator's point of view, I am  
> sensing that
> this is the proverbial can of worms, and I'm hesitant to open it.  I
> thought perhaps I could just do a quick survey, both of what people  
> do on
> their individual networks, and of what they know about others.
>
> - Do you offer QOS services across your network for VOIP or other  
> types of
> traffic?
>
> - Do you do this on a per-customer basis, or is it done globally?
>
> - If per customer, are there charges involved?
>
> - Are there any major carriers that give preference to VOIP traffic?
> I've heard rumors that Global Crossing does.  I've also heard  
> rumors that
> almost all of them do.  I've also heard that almost none do.
>
> - For those that offer QOS services for VOIP, is traffic  
> classification
> done by TCP layer protocol and port number?
>
> - It's hard for me to imagine major carriers implementing any kind of
> standardized classification.  For instance, if most carriers agree  
> that
> UDP traffic destined for ports Y through Z is VOIP and should thus be
> treated in a more time-sensitive manner, what's to prevent the newest
> version of BitTorrent to take advantage of this by fitting it's  
> traffic to
> that profile?  One of the things that makes VOIP so frustrating to the
> telcos - data is data is data - seems to bite VOIP in the ass  
> here.  Is my
> thinking correct?  Or maybe traffic classification is happening at a
> higher layer?
>
> I can only imagine this issue growing larger.  With streaming audio  
> and
> video well-established, VOIP at the dawn of major acceptance, and  
> Internet
> TV, movies, and other real-time dependent applications on the near
> horizon, it's looming large.
>
> I know aspects of this have been brought up before, notably the  
> whole "the
> Internet core can't take it" conversation.  I'm hoping my questions  
> are a
> little more specific.  (However, I do find the aforementioned threads
> quite fascinating, for the record.)
>
> Thanks,
>
> Rick Kunkel
>

-- 
Jo Rhett
senior geek

Silicon Valley Colocation
Support Phone: 408-400-0550







More information about the NANOG mailing list