Older Catalyst 40003/4006 Sup1 & Sup2 & L3 Module & QoS

[andrew.butterworth]

I have a customer with some older Catalyst 4003 chassis with Supervisor1's and some 4006 with Supervisor2's. Each of the switches has the WS-X4232-L3 services module and this is being used to connect to a 6500 Core using /30 links. From a classic networking topology of Core, Distribution & Access this fits reasonably well, although some of the access-layer is collapsed into the 4000's on 10/100 ports.

We are in the process of implementing Voice over this network (IP telephones etc) and I need to know what we can expect to achieve with the 4003 & 4006's. I have read the QoS capabilities of the Supervisor1 & 2 (CoS only with round-robin 2Q1T) and the 4232-L3 module with 4 weighted round-robin queues based on IP Precedence (or at least the first 2 bits of IP Precedence.....).

The IP Phones set CoS 5 & DSCP 44 for all IP Packets (there is no distinction between signalling and media but this can be achieved with newer switches using QoS policies to re-classify). I can engineer things so that the CoS received from the IP Phones places them into Q2 and then the IP Precedence places the traffic into Q4 when leaving the 4232-L3 module to the Core.

The Core is 6500's so I can re-classify the traffic here but this is not desirable.

Is any of this worth the bother or should the equipment be replaced?

Thanks

Andy

[pflunkert]

Hi Andy,

when you are a reseller it#s better to replace the old hardware ;-)))

But it's not necessary. When the switch trust the DSCP values it should not a problem. Normally the phone sent Cos 5 and DSCP 46, but you can place also DSCP 44 in the priotity queue in the Core. Read the qos section in the configuration guides and you see that all will works.

regards

Peter

[andrew.butterworth]

The issue I can see though is the traffic flow from the IP Phones at the Catalyst 4000 end when contending for the Uplink to the Core as this is shared with PC's. Without a strict priority queue we can only schedule traffic into either of the 2 Round-Robin queues so potentially we will have congestion and delay issues.

I agree I can trust (or reclassify) the DSCP values received on the Core Uplinks but traffic scheduled out of the 4000's is not too controllable.

Andy