[c-nsp] Cisco ToS/CoS/DiffServe

[Saku Ytti]

On (2006-03-31 23:38 -0800), Lawrence Wong wrote:

Hello,

> I read from the documentation that switches like the
> 2960 and 3560 support QoS based on ToS/CoS/DiffServe,
> etc.

prec5 (101) == dscp40 (101000) == tos160 (10100000)

 CoS is 3 bits also. prec,dscp,tos are all from IPv4
'tos byte' just giving remaining extra bits different
meanings, so changing prec changes dscp and tos too.
 CoS of course is in 802.1q headers only and can have
different value to that off 'tos byte' content.

> Does anyone have any idea how the QoS settings can be
> verified? Or how will I know whether the actual
> traffic has indeed been given the CoS which I wanted
> to set?

lab, lab, lab and lab testing if you're service provider.
catalysts aren't the most trivial platform to get QoS
right :/, using extensive lab testing will probably be
only way to get it right.
If you're 'user' of the QoS, then apparently you don't 
need the QoS if you can't know if it works or if it 
does not work (if you can't observe quality issues,
why bother fiddling with QoS).

example from my own connection, typical DSL connection
with large downstream, low upstrea. If my QoS were
suddenly to break down. I could immediately observe
it when I next time upload something, by my mp3
playing (over samba), tv-series (over http), interactive
ssh sessions (upstream latency) and download (upstream
tcp acks) having severe quality issues. In my case QoS
is simply prioritizing small packets (<200 bytes)
upstream, which will take care of all of the problems
that I can observe as a user. 

-- 
  ++ytti