What is MPLS-TP and is it relevant?

At the time when I was writing my MPLS books and developing MPLS courses for Cisco, everyone was ecstatically promoting GMPLS (Generalized MPLS) as the next unifying technology of everything, making someone so fed up with the fad that he wrote the Electricity over IP RFC.

GMPLS got implemented in high-end routers, but never really took off (at least I’ve never seen or even heard about it). Obviously the transport teams found the idea of routers requesting on-demand lambdas with IP-based protocols too hard to swallow.

Few years ago, ITU-T tried to reinvent the wheel (and ride on the MPLS brand name) and started the T-MPLS effort. After a while, IETF and ITU-T managed to sort out the paperwork (including plenty of liaison notes) and started joint MPLS-TP task group and new standardization efforts.

The first MPLS-TP RFCs have already started to appear and there are numerous drafts in the pipeline, so it’s high time to take a second look at MPLS-TP and figure out what’s new and whether it’s a useful technology. I tried to do that in the MPLS and MPLS Transport Profile (MPLS-TP): The technology differences article published by SearchTelecom.

18 comments:

  1. Hi Ivan,

    How do you thing about the OTS - with enhanced ethernet services deployment in enterprise or university? Basically ROADM with integrated ethernet switch on traditional node rooms, only two powerful routes/firewalls in datacenters with ring topology to each node rooms which will be layer 2 switched.

    Thanks,

    Schilling

    ReplyDelete
  2. It all boils down to whether statistical multiplexing is better than TDM :) Will this discussion ever end?

    ReplyDelete
  3. I saw a lot of presentation about this from several vendors on internet2 joint techs conferences, and we are seeing joint presentation from verisonwireless/tellabs/juniper to university. Just like MPLS is coming to the enterprise/university, will this ROADM w/ ethernet switching make the way to enterprise/university?

    ReplyDelete
  4. I think as long as BGP is not involved this is a good solution for carriers. Carriers don't have enough of the skill set to design, deploy, maintain BGP. Using the network mgmt plane fits well with their traditional SONET provisioning systems and work processes.

    ReplyDelete
  5. Maybe carriers should get the skills and reap the benefits?

    ReplyDelete
  6. On serious note, I wish you could provide some SONET/SDH vs MPLS-TP comparison. To me neither T-MPLS nor PBB-TE (it's less "successful" peer ;) make much sense for an operator providing circuit transportation services. Bandwidth efficiency? Come one, VCAT/LCAS make SDH circuits pretty flexible. Packet services? MPLS-TP provides PWE3 services and.. that appears to be it. How is that significantly better than TDM circuits? If a carrier is looking toward proving rich packet services then it makes more sense to run a time-proven IP/MPLS network core. Lastly, speaking of MPLS-TP OAM-P features designed similar to SONET/SDH standards: if we are making PSN that similar to TDM, maybe there is no reason to bother with it and stick with TDM in the first place?

    ReplyDelete
  7. MPLS-TP mainly will use in the access network because in the upcoming years it will become more difficult to manage the nodeBs and enodeBs network.

    ReplyDelete
  8. Infinera is using GMPLS in their gear - and it works swimmingly well for the control plane. I think your argument may be off basis slightly.

    ReplyDelete
  9. That's fantastic. Thanks for the update. A few months ago when I started looking at MPLS-TP I tried to find out if any vendor supports GMPLS and the only hits I got were Cisco and Juniper routers.

    Would you know how many transport networks actually use GMPLS versus more traditional NMS-based provisioning?

    ReplyDelete
  10. Infinera is all based on GMPLS for the control plane - in fact you can't implement it without. I am under the impression that Infinera is in most large carriers and web services at this point. Their PIC technology outperforms everything else in the optical market right now. I personally am using the equipment at a relatively large MSO and it's, hands down, the best platform out there. All of our Nortel gear is being replaced with it.

    ReplyDelete
  11. Peter PalĂșch28 July, 2011 14:07

    Petr,

    Although this is a reply that comes more than half a year late, I hope you will still get to read it :)

    I believe that the point of MPLS-TP is having a telco-kind of management and control above the network. From the service point of view, there is no difference to MPLS or MPLS-TP or T-MPLS... every one of them uses the label stack precisely in the same way. In my opinion, the difference is how do the labels (and LSPs and all the stuff) get set up in the first place, and intelligence on LSRs is needed for that.

    We already have quite a lot of label-disseminating protocols: LDP, RSVP, BGP, the first two of them being heavily dependent on IGP routing protocols. Providers routinely dislike running RSVP in their networks, even if for sole purposes of MPLS-TE. So in order to actually set up a LSP, there is a lot of work to do in the control plane of a router, and synchronizing this work among lots of routers is a nontrivial task.

    The other observation is that a LSR actually does not need IP for most of its data plane operations, as it is concerned with labeled packets most of the time. All the hassle with IP-based protocols is just to create the label bindings - and then, you can theoretically get rid of the IP stuff as long as you have your LFIB populated with correct entries.

    In telco network, the idea of a management network above the managed nodes, and the managed nodes being "fast and stupid", has its advantages. A central node managing the network allows for complex and extensive path creation and selection between nodes because it is performed as a centralized computation, as opposed to all nodes in the network performing this computation in a distributed manner, without having a centralized point of command. This is where I see the MPLS-TP step in: as a means of controlling the LSP creation and management and removing the IP-based control plane from the network nodes and centralizing it in a single control element who is in charge of all network nodes. For example, the MPLS-TE is a complicated thing to do if lots of tunnels are to be configured in a regular MPLS network. The MPLS-TP could solve this quite easily, as all the hassle of OSPF-TE or ISIS-TE combined with RSVP-TE would simply not be necessary.

    I also think that nobody is trying to reinvent the TDM or SDH by MPLS-TP. Surely, the underlying technology under MPLS is mostly packet-based, so it is almost impossible inherently to maintain the time division aspect of any TDM/SDH network. What we are getting at with MPLS-TP is having more controlled environment where the LSPs are built not by cooperation of network nodes like in today's MPLS networks, but rather by a centralized control. Reminds me slightly of LWAP/CAPWAP (although they centralize both data and control plane).

    Best regards,
    Peter

    ReplyDelete
  12. I recently started looking at MPLS-TP as part of adding OAM support for the same. Wondering just what's different from creating plain old static LSPs (say by this method on cisco: http://www.cisco.com/en/US/docs/ios/mpls/command/reference/mp_m2.html#wp1060671)

    Granted, MPLS-TP also says you can use GMPLS but the 'static' way seems to be already there...? So what's new here ?

    ReplyDelete
  13. Correct Cisco link:http://www.cisco.com/en/US/docs/ios/mpls/command/reference/mp_m2.html#wp1060671

    ReplyDelete
  14. does mpls-tp still uses ethernet interfaces with pwe or do you need to have pos/sdh interfaces to implement mpls-tp?

    ReplyDelete
    Replies
    1. Would have to go through the standards for a definitive answer, but from the purely conceptual perspective the L2 protocol should not matter. You can use MPLS labels across Ethernet or POS. Individual vendors or products might have specific restrictions, so check with your vendor.

      Delete
  15. Hi There,

    I am planning to do my masters project in MPLS-TP. Can someone give me some idea and guide me through...???

    Thanks

    ReplyDelete
    Replies
    1. Sure. That's the job of your mentor/thesis advisor.

      Delete
  16. We know MPLS-TP is l2 VLAN based .we can add Maximum 2047 Node due to VLAN Limitation for MCC and SCC VLAN. Any one could give me answer how to manage more than 2047 LTE Node in MPLS-TP network.

    ReplyDelete

You don't have to log in to post a comment, but please do provide your real name/URL. Anonymous comments might get deleted.

Ivan Pepelnjak, CCIE#1354, is the chief technology advisor for NIL Data Communications. He has been designing and implementing large-scale data communications networks as well as teaching and writing books about advanced technologies since 1990. See his full profile, contact him or follow @ioshints on Twitter.