View Only
last person joined: yesterday 

Ask questions and share experiences about ACX Series, CTP Series, MX Series, PTX Series, SSR Series, JRR Series, and all things routing, including portfolios and protocols.
  • 1.  Correct CoS field for MPLS

    Posted 06-18-2016 11:31



    I am studying CoS at the minute and wondered if anyone could help clarify something for me.


    I want to put my network control traffic into an LSP which I can then traffic engineer to a high level of protection and low latency path


    I want to apply CoS to the network control traffic


    So I have an output filter set on Lo0 to apply CoS to any traffic leaving Lo0 towards the range of addresses my other core node loopbacks are on:




    term ldp {
    from {
    destination-address {;
    protocol [ tcp udp ];
    port ldp;
    then {
    forwarding-class network-control;
    dscp 0x38;




    My question is: Given that I am putting this traffic through an LSP, in order to get the next node in the path to honour the CoS, should I be setting this to EXP instead of DSCP?

    I would assume the next node will be looking at the MPLS header, expecting EXP to be defined and, upon not seeing a defined EXP, it will ignore the DSCP I have set.


    Is this the case?

  • 2.  RE: Correct CoS field for MPLS
    Best Answer

    Posted 06-20-2016 19:23



    An LSR would only look at the MPLS header to label-switch the packet, thus any CoS required needs MPLS EXP bits to be set.

    As part of the CoS Configuration, you can configure and apply rewrite rules to mark/re-mark the packets accordingly on the egress interface such that Core facing MPLS interfaces gets MPLS EXP bits set based on the forwarding class the packet entered [either through BA or MF classifier].



    class-of-service {
    	rewrite-rules {
    		exp CORE_EXP_REWRITE {
    			forwarding-class NC {
    				loss-priority low code-point 110;
    	interfaces x/x/x {
    		unit * {
    			classifiers {
    			rewrite-rules {
    				exp CORE_EXP_REWRITE;

    When using MF classifier at the network edge [for example ingress PE], setting a DSCP marking would mark the IP header which may be left intact going through an MPLS core and be read and CoS treated on the egress PE which would look at the IP header.

    Setting the forwarding-class would queue the packets and may be sufficient along with rewrite rules to EXP mark the MPLS header on MPLS core egress interfaces.


    I also recommend as good practice to set the loss-priority in the MF classifier to avoid any mistreatment on the CoS path.


    Worth noting also is that routing protocol traffic is marked with DSCP48 [ToS 0xc0] by default.

    By wanting to put ldp traffic into an LSP, I guess you are referring to tLDP.