Routing

last person joined: 2 days ago 

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.  QoS for voice from MPLS CE to CE

    Posted 07-28-2013 03:38

    I would like to implement end-to-end QoS for customer's voice with EF. The topology is quite simple EX4200-CE1-->MX960-PE1-->MX960-PE2-->EX4200-CE2. I know that QoS configuration is unidirectional , so that I have to configure bi-directional between CE1 to CE2 . Becoz both of normal IP traffic and MPLS traffic are running between PE1 and PE2. Then now,  two issues:

    a. Should I rewrite default forwarding class to mpls exp expedited-forwarding on PE1 outgoing interface and rewrite mpls exp expedited-forwarding to default forwarding class expedited-forwarding on PE2 outgoing interface ?

    b. How to treat on normal non-mpls traffic?

     

    Thanks for your comment if any.



  • 2.  RE: QoS for voice from MPLS CE to CE

    Posted 07-28-2013 07:06
    @Hugo Chen wrote:

     

    >  The topology is quite simple EX4200-CE1-->MX960-PE1-->MX960-PE2-->EX4200-CE2.

     

    I would assume this is a Layer-3 VPN; is that correct?

     

    > a. Should I rewrite default forwarding class to mpls exp expedited-forwarding on PE1 outgoing interface

     

    Yes. You should classify traffic from CEs (how, it is up to you) so that voice is put in the EF class, and rewrite on the core-facing interfaces with an EXP rewrite rule.

    Since traffic between PEs is MPLS you need to make sure that traffic coming from your core-facing interfaces is correctly classified.


    > and rewrite mpls exp expedited-forwarding to default forwarding class expedited-forwarding on PE2 outgoing interface ?

     

    No, you don't need to have any rewrite rule on CE-facing interfaces, unless you need it for some other purpose.

     

    > b. How to treat on normal non-mpls traffic?

     

    It should not make any difference, as you do not have any classifier anywhere that would put traffic in your EF class, except for your voice traffic from your CEs.
    However, I would advise to do a complete, extensible CoS design instead of just enabling one additional class for voice. This is non-trivial, but will probably save you problems later on.
    Hope it helps,
     Saverio


  • 3.  RE: QoS for voice from MPLS CE to CE

    Posted 07-28-2013 07:34

    Hi Saverio

        Thanks for you reply, really appreciate. It's my first time to configure CoS on Junos device , feel a little bit confusion. Have some more few questions need to be clarified.

     

    1、What's difference between VPLS and L3VPN in QoS. Becoz we are using the same IP header, so I can't figure out what's the difference.

     

    2、Have no idea why PE2 interface that facing to CE2  no need to rewrite default forwarding class expedited-forwarding. Becoz CE2 don't have any Queue information after MPLS label pop-out. Just can't figure out.

     

    Thanks in advanced



  • 4.  RE: QoS for voice from MPLS CE to CE
    Best Answer

    Posted 07-28-2013 07:44

    Hi Hugo,

     

    1、What's difference between VPLS and L3VPN in QoS. Becoz we are using the same IP header, so I can't figure out what's the difference.

     

    If it is VPLS, it is the same - the only difference is classification, i.e. how do you recognize voice traffic - other than this it is the same.

     

    2、Have no idea why PE2 interface that facing to CE2  no need to rewrite default forwarding class expedited-forwarding. Becoz CE2 don't have any Queue information after MPLS label pop-out. Just can't figure out.

     

    This is purely dependent on how you design you CoS marking on your Layer-2 domain.

     

    For example, if you are using Vlans, you can consider using an 802.1p classifier that looks at the PCP bits within the Vlan-tagged frames; those bits should be preserved from one CE to the other, so you should not need to rewrite them on egress.

    It is personal preference, but that is what I would do - I would not want to configure layer-3 classifiers on layer-2 interfaces (even if it can be done - on most platforms).

     

    Saverio