[afnog] QoS/TE

[Mark Tinka]

On Wednesday 18 January 2006 22:57, Brian Candler wrote:

> 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.

If core links are running at 20% utilization, including 
the occasional spike in traffic that could send that up 
to 60% or more for a couple of minutes to a few hours 
(school-kid gamers back home from school, DoS attacks, 
breaking news, e.t.c.), with enough bandwidth on the 
wire, why would you expect an increase in latency if your 
gear can run the wire at line rate, and you have a good 
mind to keep utilization under a rough figure, e.g., 75%?

> 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.

So then this begs the question; (I know it's not an exact 
science, and each situation is unique) specifically for 
delay-sensitive traffic like VoIP and video:

a) At what threshold do you determine deploying QoS/TE in
   the core, regardless of whether you've reached line
   rate capacity or not?

b) At what threshold do you determine upgrading raw core
   media/bandwidth to maintain good performance of your
   (non-)QoS-enabled core?

c) At what point do you determine QoS may be actually
   increasing latency especially on software-based
   routers?

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: not available
Url : http://listserv2.cfi.co.ug/pipermail/afnog/attachments/20060119/384436b0/attachment.bin