Re: [mpls] draft-kini-mpls-entropy-label-src-stacked-tunnels

["Nobo Akiya (nobo)" <[e-mail]>]

Hi Stephane, et al,

Adding authors of draft-ravisingh-mpls-el-for-seamless-mpls as I believe there's benefit in creating a consistent solution.

I strongly believe (and hope) that we do not want to go with a solution where label stack has multiple ELI/EL at any point in time. Increase in label stack size is concerning, but multiple ELI/EL becomes more concerning from OAM perspective, i.e. it becomes very difficult to compute EL to traverse specific path. Thus, I also think solution 3.3 described in draft-kini-mpls-entropy-label-src-stacked-tunnels is the best idea. But there is one issue.

Q: How does LSR, popping the outer label, determine whether it needs to carry over ELI/EL (re-insert ELI/EL below exposed top label)?

Section 5.3.2 of draft-ravisingh-mpls-el-for-seamless-mpls states:

  At an ingress LER, the router SHOULD not insert an (ELI+EL) for an
   LSP if the packet already contains an ELI.
   This ensures that for a data packet on a hierarchy of LSPs, there
   will be only 1 instance of an (ELI+EL). This helps to prevent the
   issue of section 4.3.1.

Taking the example from draft-ravisingh-mpls-el-for-seamless-mpls ...

                S1                  D1
                  \    ---------    /
                   A===B===C===D===E
                  /    ---------    \
                S2                  D2

   In the above topology, let there be the following LSPs:

        L1: B->D
        L2: A->E, tunneled through LSP L1
        L3: S1->D1, tunneled through LSP L2
        L4: S2->D2, tunneled through LSP L2

Let's say:

(1)    S1 does not push ELI/EL.

(2)    A pushes ELI/EL.

(3)    B does not push ELI/EL, since label stack already contains ELI/EL.

Behavior so far aligns with 5.3.2 of draft-ravisingh-mpls-el-for-seamless-mpls (snippet above). Ideally what should happen is:


(4)    D pops ELI/EL, and pushes (or carries over) ELI/EL below exposed top label.

(5)    E pops ELI/EL, and _does not_ push ELI/EL below exposed top label.

(6)    D1 receives data without any ELI/EL (as expected).

How does nodes D and E determine the right behavior?
>From what I've read in the two drafts, I don't see how this behavior is determined.

If we can address this issue, I think solution is applicable to Segment Routing label stack with ELI/EL, meaning behavior can apply to solution 3.3 of draft-kini-mpls-entropy-label-src-stacked-tunnels.

One possibility is to make use of TC or TTL of EL to keep track of carry-over number. Meaning:

When ELI/EL is not inserted in a new tunnel because ELI/EL exists in the label stack, then:

-        Increment carry-over number.

-        Move ELI/EL to below top label.
When data terminates a tunnel, then:

-        Decrement carry-over number.

-        If (carry-over != 0) re-insert ELI/EL to below exposed top label.

I'm no hardware expert either, and not sure if something like above can be implemented. But if possible, then ELI/EL pushed in SR network can pre-set carry-over number to certain value to ensure it is carried over below top label all the way though, and there is only one ELI/EL at any given time.

All this to say, I think there's a benefit in discussing this topic further, with both drafts in mind.

-Nobo

From: mpls [mailto:[e-mail]] On Behalf Of [e-mail]
Sent: Friday, January 31, 2014 9:12 AM
To: [e-mail]
Cc: [e-mail]
Subject: [mpls] draft-kini-mpls-entropy-label-src-stacked-tunnels

Hi Authors,


I would like to know if you are progressing on this draft ?
Entropy label support for stacked tunnels would be mandatory for us, so I would like to support the work on this topic to have a working solution as soon as possible.


Regarding the solutions you are proposing in the current version, there is none perfect one, and unfortunately all have drawbacks.
The compatibility (on LSR) with current generation of hardwares may be "good" (mandatory ?) for the target solution.

In your document , solution §3.3 "re-usable EL for a stack of tunnels" sounds for me to be the best base idea. I'm just wondering what would be the hardware impact of doing the reinsertion of ELI at tunnel end . Could this be done in one pass ? (IMHO, this may be possible, as today we are able to pop or swap + push FRR headers) As you are three different vendors as co-author, did you already evaluate such impact on your hardwares ? (I'm not expecting details on the mailing list, but I would be interested by details unicast to me and just yes/no on the the list)

To be exhaustive in listing solutions, did you think about leaving the EL/ELI at top of the stack ? I think there is already a case where a special label may be kept at top of the stack (MPLS Router alert).
What would be needed :

-        each hop need to advertise is ability to process EL, if nexthop cannot process EL, it should be removed when forwarded to nexthop (there should be the same requirement for re-usable EL)
I don't think this is really different from re-usable EL in the concept :

-        reusable EL : process top level forwarding label, if popped and next label is EL, pop ELI/EL and next label L, push pack ELI/EL and then L (we need to swap positions between ELI/EL and L) if nexthop is able to process EL.

-        top level EL : process top level ELI, ELI is recognized, ELI/EL is removed, forwarding is done on forwarding label, ELI/EL is pushed back in nexthop is able to process EL.

In term of operations, I think that top level EL may be simpler , but as I'm not hardware coder, may be I'm wrong ... moreover it may be similar to MPLS Router alert processing.


Your thoughts ?


Stephane




_________________________________________________________________________________________________________________________



Ce message et ses pieces jointes peuvent contenir des informations confidentielles ou privilegiees et ne doivent donc

pas etre diffuses, exploites ou copies sans autorisation. Si vous avez recu ce message par erreur, veuillez le signaler

a l'expediteur et le detruire ainsi que les pieces jointes. Les messages electroniques etant susceptibles d'alteration,

Orange decline toute responsabilite si ce message a ete altere, deforme ou falsifie. Merci.



This message and its attachments may contain confidential or privileged information that may be protected by law;

they should not be distributed, used or copied without authorisation.

If you have received this email in error, please notify the sender and delete this message and its attachments.

As emails may be altered, Orange is not liable for messages that have been modified, changed or falsified.

Thank you.