Qos policy

Djule2804 · ‎06-11-2007

I must build a QoS policy on network having different sort of trafic such as real time trafic, best effort trafic...

My test plan is formed by LAN1 which is linked by SWITCH1 to ROUTER1.

ROUTEUR1 is linked to ROUTEUR2 and ROUTER2 is linked to LAN2 by SWITCH2.

It should have congestion point on link bounding ROUTER1 and ROUTER2 (wan).

My question is : Should I shape or police trafic on switchs to avoid to create congestion point on wan link?

What is the difference between use of policing, shaping and use of congestion avoidance mechanism such as WRED or RED?

Which mechanisms are used in pratice and on which sort of traffic?

paolo bevilacqua · ‎06-11-2007

Hi,

the best way to learn these things is by reading the whitepapers on the cisco website. There are many and they will answers all your questions.

Anyway in short, if the router with wan interface is configured correctly, there is no need that you do anything on the switches.

The mechanism that I suggest in practice is a strict priority (LLQ) for voice traffic on wan links, and nothing else.

Hope this helps, please rate post if it does!

Djule2804 · ‎06-11-2007

Thanks for your response.

Actually, i'm studing all the white paper published by cisco to elaborate my strategy and i thought to apply a QoS policy on the switch to have a flow-based policy with finer granularity than on the routeur.

If i drop real time trafic excess, all real time trafic flows can be affected whereas the excess would come from a single flow.

So, it may be more judicious to drop excess trafic coming from this single flow. Is it possible to build this policy on the switch?

paolo bevilacqua · ‎06-11-2007

Hi,

In practice, the switch does not see congestion and will not drop traffic. The QoS configuration must be done on the point of congestion, that is WAN interface on the router.

Djule2804 · ‎06-11-2007

Thanks for your response.