Hi,

Using wfq (ie class based shaping) : 

To illustrate, lets suppose the following weight for each queue, a:1 b:2 c:3

If total bandwith doesn't exceed 45Mbps, no congestion: each queue can manage 
its bandwith.

If total amount of bandwith 45Mbps exceed, wfq gives a weight to each queue, 

In your case, if only 10Mbps is sent through queue c:, mean 35M left, following 
scenarios exist:

a: AND b: in congestion  will resp share the left 35Mbps :

a: 35*1/(2+1) = 12M
b: 35*2/3 = 23M
c: 10M

a: OR b: in congestion, wfq will give 

a: 45*1/(1+2+3) = 7.5M
b: 45*2/6 = 15M
c: 45*3/6 = 22.5M

In other scenario, where a: AND b: AND c: in congestion,

a: 45*1/6 = 7.5M 
b: 45*2/6 = 15M 
c: 45*3/6 = 22.5M


wfq is a basic way to share bandwith using the wfq algorithm.
Cisco MQC manage it very well and you won't need to manage weight as done in 
the example, just give the bandwith to queue and the max bandwith as a 
reference bandwith for the link...

You can associate with this mechanism a more complex one which is wred to avoid 
congestion for tcp flow. But it is another history... 

To monitor with snmp, need to use the Cisco CBQoS Mib ... 

HiH and HnY ...
Rachid

-----Message d'origine-----
De : [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] De la part de Rivo Tahina 
RAZAFINDRATSIFA
Envoyé : mardi 8 janvier 2008 11:39
À : [email protected]
Objet : [c-nsp] traffic shaping on 7200

Hi all,

I do traffic shaping on a 7206 box, eg:
a: 10Mbps for one class
b: 15Mbps for another class
c: 20Mbps for another class

If c: only use 10Mbps, how can I share it to a: and b:?
How can I monitor (graph) the bandwidth usage for the 3 classes?

Best Regards.

_______________________________________________
cisco-nsp mailing list  [email protected]
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/
_______________________________________________
cisco-nsp mailing list  [email protected]
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

Reply via email to