CIR and congestion

[mathieuploton]

It is not exactly a cisco topic related but I am unable to find a forum for this brand so I thought cisco guys should be the best to answer a QOS question :-)

I have a wimax network with qos capacities enabling each customer to have a guaranteed and a max bandwitdh (upstream and downstream). This provides two parameters to each customer a CIR and a MIR (which is actually the CIR+ Burst).

The sum of CIR exceed my device capacity which is wrong I know, that means the bandwitdh is not really garanteed to each customer. I was just wondering what happens in the case of congestion. How the device is handling all those CIR exceeding his capacity ? And what will be the difference if I decrease my CIR to be lower than the capacity and I am still in this hypothesis of congestion ?

Basically, if my device capacity is 2 Mbits :

What will be the difference if I have
- 4 users with a CIR of 1Mbits
- 4 users with a MIR of 1Mbits
and they try using all of it ?

[Laurent Aubert]

HI,

It's like not having QoS at all ;-). Depending on customer application behavior and which packets are dropped by the device, you could end with any scenarios like:

- 4x customers using each 500Kbps (perfect world)

- 2x customers using each 1Mbps and 2x customers starving

- Anything in the middle.

HTH

Laurent.

[mathieuploton]

Actually there is a graceful degradation mode which equally redistribute the bandwidth to customers.

But I do understand that this is not the solution and I should certainly switch off the CIR and use burst instead.