I would not recommend using Uniform mode as it modifies Customer(CE) markings With Pipe mode the scheduling egress from MPLS cloud towards the CE is based on MPLS EXP markings -that will allow you use only generic outbound policy With Short-Pipe mode the scheduling egress from MPLS cloud towards the CE is based on CE markings -that will allow you to use granular outbound policy profiles per customer contract The practice is to classify, mark and condition the traffic ingress to MPLS cloud on PE based on the contract policy profile If you manage the CE you can classify, mark and condition the traffic on CE (by conditioning I mean shaping, policing and congestion avoidance) On the MPLS links you should base you PHB on MPLS EXP and you should not need any conditioning -as the core link capacities should be sufficient Usually only couple of classes are used in the core like: Priority, Conform and Un-conform
adam -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Saku Ytti Sent: Wednesday, July 11, 2012 12:19 PM To: [email protected] Subject: Re: [c-nsp] ASR9000 MPLS QOS Issue On (2012-07-11 18:06 +0800), Xu Hu wrote: > Yes, you are right, now my big concern is that if one packet have one DSCP value, then after transfer through the MPLS network (deploy the MPLS EXP QOS in the MPLS network) the output packet will attach the same DSCP value or not? Both are supported. Uniform mode where it is DSCP -> EXP, EXP -> DSCP. And Pipe mode where in ingress you set EXP was you wish and in egress you can schedule based on EXP, without writing it in DSCP. I.e. Pipe mode retains original DSCP, not mangling it, not necessarily even using it. Uniform mode transports DSCP as EXP. I personally would always do Pipe mode as Core QoS and Access QoS are two different things. In Pipe mode depending on platform, you store EXP in internal DSCP or copy it via QoS config to 'qos-group' in MPSL2IP and then in IP egress, you schedule based on qos-group value. ASR9k does support qos-group, so I would suggest using them. -- ++ytti _______________________________________________ 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/
