[afnog] QoS/TE

[Brian Candler]

On Wed, Jan 18, 2006 at 11:19:40AM -0800, Bill Woodcock wrote:
>       On Wed, 18 Jan 2006, Jim Forster wrote:
>     >>> Is QoS/TE in the core really necessary.
>     >> Considering the capacity of most modern core equipment,
>     >> I tend to say avoid QoS there.  It's usually cheaper
>     >> to upgrade the core than to deploy QoS.
>     > I would tend to agree, although most of the thought and analysis has  
>     > been done for places with fiber based cores.   In these places the  
>     > cost of additional bandwidth for core networks is generally cheaper  
>     > than additional bandwidth for access networks.

FWIW, I know a large UK ISP which *has* deployed QoS in the core -
specifically, low-latency queuing with MPLS. It's a simple model, where
there are low-latency queues for delay-sensitive applications like VoIP,
and then there's everything else.

When discussing this I made the same points as have been raised here. The
response was that even relatively uncongested links, around 20% utilisation
upwards, can result in measurable increases in latency and jitter. So if you
have a network link with moderate utilisation (say 50%) and a relatively low
proportion of high-priority traffic (say 20% of that), the performance for
the high priority traffic can be made noticeably better.

There's another reason for implementing QoS in the core, and that's purely
as a marketing gimmick. Some customers will associate "quality of service"
with "good quality service", or at least recognise it as a technical
buzzword which your competitor doesn't have. If rolling out QoS means more
people paying you money, then hey, why not :-)

Regards,

Brian.