From: Serhat Aslan (serhatworks@gmail.com)
Date: Mon Aug 06 2007 - 20:52:37 ART
Hi Toh,
I couldn't traced all the statement at the questions, (but I tried :)) and
I add some comments below:
Proposed process: ingress(!) (dscp-mutation) --xxxx---(wrr-queue)(egress),
switch ingress operations operations are very limited due to performance
limitations, so packet classifications are made by the dscp values, then
they manipulate these values for its egress operations.
Mapping :64 dscp -> 8 Cos - > we have got only 4 queue :) (*) (summary
of the problem :))
methods: ingress->egress
dscp-map -> wrr-queue
cos-map -> wrr-queue
((dscp -> dscp) == mutation) -> dscp-map -> wrr-queue
Dscp-mutation : logical equivalent to a a DSCP-to-DSCP mapping. Ingress
DSCP mutation is commonly used when connecting multiple autonomous QoS
domains (**)
If you can, concentrate the problem as considering limited hardware
sources for identification and classification the packets for satisfy
end-to-end QoS operations (voip,etc..), I think you will better understand
the transition of the packets.
*-> Statement taken from GS archive, Scott Morris letter,
** -> Explanation taken from "Cisco Catalyst QoS" book.
---------
Serhat Aslan
On 8/7/07, Toh Soon, Lim <tohsoon28@gmail.com> wrote:
>
> Hi Herbert,
>
> So, the mapping is done based on DSCP values at the very beginning when
> the
> packet enters the switch, to determine its WRED threshold when it later
> exits the switch? I was under the impression (from theory in books) that
> we
> choose the thresholds based on CoS (is this referring to the CoS that's
> mapped to from the internal DSCP?), though I can't find any command to do
> that. Secondly, what if I configure "mls qos trust ip-prec" on the ingress
> int, can we still perform the DSCP-to-threshold mapping?
>
> Another thing that baffles me is mapping the internal DSCP values to CoS
> values for egress queueing/scheduling, where we can change the default
> mapping using the global config command "mls qos map dscp-cos". What if
> the
> packet exits a non-dot1q link, how is this CoS value carried in the
> packet?
> Or is that CoS value just an internal value to the switch?
>
> Appreciate any comments and guidance.
>
>
> Thank you.
>
> B.Rgds,
> Lim TS
>
>
> On 8/7/07, Herbert Maosa <asawilunda@googlemail.com> wrote:
> >
> > I believe this is used for both WRED and Tail Drop, to specify the
> mapping
> > between DSCP values and drop thresholds.
> >
> > For each of the four queues, you specify two drop thresholds ( low, high
> > ). Then you need to specify which DSCP values fall under either one of
> the
> > two thresholds. This is where you use the wrr-queue dscp-map command, to
> map
> > the DSCP values to the thresholds.
> >
> > Herbert.
> >
> >
> > On 8/6/07, Toh Soon, Lim <tohsoon28@gmail.com> wrote:
> >
> > > Hi All,
> > >
> > > Does anyone know exactly how the "wrr-queue dscp-map" interface config
> > > command is used on a Catalyst 3550?
> > >
> > > From DocCD I understand it's used on the ingress interface to map
> > > trusted
> > > DSCP values to WRED thresholds (1 or 2). I mean, this happens before
> > > DSCP
> > > mutation (if one is configured) takes action?
> > >
> > > If DSCP mutation is configured, regardless of what the "mutated"
> > > internal
> > > DSCP value is and the resultant mapped COS value is before egress
> > > scheduling, the packets will be assigned to the WRED threshold which
> we
> > > specified earlier on the ingress interface?
> > >
> > > Please help me understand better.
> > >
> > >
> > > Thank you.
> > >
> > > B.Rgds,
> > > Lim TS
> > >
> > >
> _______________________________________________________________________
> > > Subscription information may be found at:
> > > http://www.groupstudy.com/list/CCIELab.html
> > >
> >
> >
> >
> > --
> > Kindest regards,
> > hm
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Sat Sep 01 2007 - 11:32:09 ART