Cisco Support Community
Community Member

EXP propagation


Can anyone shed some light on the issue of EXP field propagation as labels are removed or added?


1. I understand that at ingress of MPLS cloud you would use class-based policy to apply the initial EXP value based on some class-map matching by applying service-policy in either inbound direction in case you want to do on the inbound interface or in the outbound direction if you want to do on the outbound interface (probably frequently used in MPLS DS-TE), however my question is how will this field travel along the LSP?

2. Do you have to make sure that you apply proper EXP value at any point in LSP when label is being stripped off or applied or is it automatically copies between the labels?

3. Also, at egree, do you usually match on EXP value to apply proper QoS, like LLC. For example,

class-map exp5

match mpls experimental topmost 5

policy-map premium

class exp5

priority 512

interface fastethernet0/0

service-policy output premium

4. Cisco documentation speaks of three modes: Uniform, Pipe and Semi-pipe. Are these simply concept of implementations or are they really configuration parameters?

Your help would be appreciated.




Re: EXP propagation


1) the default for a PE is to copy IP precedence to MPLS exp bits in all imposed labels. With a policer you can choose to set exp to different values.

2) the popped label exp bits are not copied to next label automatically. Assuming however that you set all exp bits to the same value during imposition this should not impose a problem.

3) This would not work as the output traffic is (presumably) IP. so the trick is:

class-map exp1

match mpls exp topmost 1

policy-map exp-mapping

class exp1

set qos-group 1

class-map qos-group

match os-group 1

policy-map customer

class qos-group

bandwidth 128

interface FastEthernet0/1

description from core

service-policy input exp-mapping

interface Serial1/0

description to CE

service-policy output customer

4) The qos models are concepts. Uniform mode means you can (and will) rewrite DSCP as the settings throughout IP/MPLS are using a uniform classification scheme. Pipe and short pipe model let you transport DSCP unchanged through an MPLS domain by setting exp bits ONLY and not rewriting DSCP in any place. This is used in case the customer has different settings than the SP and does not want them to be modified.

Hope this helps


CreatePlease to create content