MPLS Label Distribution

[snarayanaraju]

There are 2 ways to distribute the lables:

METHOD 1. Piggy back the labels on the existing protocols.
METHOD 2. Running separate Protocols for Lable Distribution

Either I can use EIGRP or OSPF in First Method:

I read that EIGRP can be used with out changing the protocols becasue Eigrp does it. The implementation for distance vector routing protocols (such as EIGRP) is straightforward, because each router originates a prefix from its routing table. The router then just binds a label to that prefix.

But whereas OSPF, Each router originates link state updates that are then forwarded unchanged by all routers inside one area. The problem is that for
MPLS to work, each router needs to distribute a label for each IGP prefix even the routers that are not originators of that prefix I am in impression that OSPF orginate the routes based on the LSAs and EIGRP learns from neighbour router.

I am not able to understand contradiction.Can you give your valuable time to explain this

regards,

sairam

[Peter Paluch · Accepted solution]

Hello Sairam,

It is nice to hear from you again.

I believe that you have somewhat incorrect information regarding the MPLS label distribution process. As far as I know, no IGP protocol in fact distributes MPLS labels. You run the IGP to have your routing table contain reachable networks withing your topology. Afterwards, each router assigns labels to routes in its routing table and advertises them to its neighbors using Label Distribution Protocol. The labels themselves are not included in advertisements of any IGP, be it RIP, (E)IGRP, OSPF or IS-IS. The Method 1 as you have indicated it is not an option for IGP protocols. In a basic MPLS deployment, you run the LDP together with an IGP routing protocol - the IGP protocol itself does not propagate label bindings.

The only routing protocol that is currently able to carry labels along with the destination prefixes is the BGP.

Best regards,

Peter

[marikakis]

Hello Sairam,

I believe you are reading the 'MPLS Fundamentals' book from cisco press. Although Peter is correct about current practice, the book at this point is trying to give insight as to why IGPs have not been modified to carry the labels. EIGRP distributes routes, so can easily bind label to a prefix. Link-state protocols (OSPF, IS-IS) do not distribute routes, but rather status of links (typically not changing originator's information) and later make independent decisions about the routes to install in their own routing table. They don't distribute their actual routes, so that's what the text means by saying those protocols would "need to be enhanced in an intrusive way to be able to do this' (i.e. distribute labels for prefixes, even those who are not originators of a prefix). Basically, this is a hypothetical scenario that wasn't chosen by implementers for the reasons described in the section you are reading.

Kind Regards,

Maria

[Peter Paluch]

Hello Maria,

As always, you are correct Enhancing link-state routing protocols with label-distributing features is possible but in most cases, it is not worth the effort. If we were speaking about OSPF, a router would need to emit LSAs with a link-local scope that would contain a list of routes in the router's routing table and the corresponding labels. With IS-IS, well, I must admit that I do not know if there is anything like TLV with a link-local scope (although there should not be a problem with implementing one). In any case, having a link-state routing protocol independently convey a list of networks and corresponding labels (which is almost the same as running a distance-vector) is somewhat of a duplication of effort.

Best regards,

Peter

[marikakis]

Hi Peter,

I am not always correct, so beware! By posting in public one puts their understanding of things to the test, so as years go by the more correct they are after been corrected many times.   To be honest I haven't put much thought in turning IGPs into label distribution protocols. My general opinion about implementation choices is this: At an early stage modularity is preferred (i.e. separate protocols/software modules) to capture abstractions, because it speeds up/simplifies development while avoiding bugs (imagine OSPF bugs combined with OSPF label distribution bugs, OSPF is complicated in itself). As more experience is gained from the implementation and everything starts looking 'straightforward', people start complaining about speed. For example, they could say things like: how about sending labels together with link updates to speed up convergence?

Kind Regards,

Maria

[Peter Paluch]

Hi Maria,

First of all, congratulations to your gold star

Yes, you are right. I am myself not a proponent of everything-encompassing-and-all-doing-protocol approach while slapping two distinct things together as route discovery and label mapping distribution means doing just that. Sometimes, integrating things can be good for various reasons but as you have pointed out and I agree wholeheartedly, it also brings a new field of problems - not just from the added functionality but also from its interaction with the existing functionality.

But - well, that's the difference between the academic and the engineering approach

Best regards,

Peter

[snarayanaraju]

Hi Peter & Maria , !!!!Wish you merry christmas!!!

It is indeed good to see your response and bit lucky that you replied to my fundamental doubt. Thanks Peter, I am interacting with u after a long gap.

As Maria rightly pointed out I am reading "MPLS Fundamentals" text book.

The intention of this post is not to do research on how to distribute labels using IGP, But to know why EIGRP does this whereas OSPF donot. After reading your valuable comments and reading the text again, I got the hint.

let me try to briefly describe: OSPF LSA database contains the prefix with orginator as the original router (with in an area). EIGRP advertise the routes as it as the next hop to reach the prefix. So as per LDP rule EIGRP is perfect tool.

Hope I am right in the understanding.

I am not able to find the RATING menu after this web page is modified. Please let me know how to rate your postings then.

sairam

[marikakis]

Hi Sairam,

None of the IGPs is actually used for label distribution (including EIGRP). If implementers wanted to make them work for label distribution, then the work would be easier for EIGRP compared to link state protocols. The reason is this: You need to associate a label with a prefix in routing table before passing the label to another router. EIGRP distributes routes, so can easily add allocated label to route sent to neighbors. OSPF doesn't send routing table prefixes to neighbors. It rather sends status of links. So, as Peter already said, OSPF would have to function both like a link state and a distance vector protocol (i.e. send both link state updates and routing table entries to neighbors).

Sairam, if you proceed with the rest of the book and read more about how things are done in practice, I believe details will be more clear. Sometimes introductory chapters in books are more difficult than the rest of the book, because the authors try to give a general picture, and mention sometimes concepts that actually need you to read (perhaps) the entire book to fully understand them. I usually go back and read again introductions for this reason.

Kind Regards,

Maria

p.s.1 Peter, thanks for the comments about my badge

p.s.2 Merry Christmas everyone! With the gold badges in this thread at this time of year, we made ourselves a quite appropriate Christmas environment

p.s.3 You use the stars on bottom left of post to rate posts. Careful because it happens quickly, and if you accidentally give us 1-ratings we might not respond again to your questions!

[Mohamed Sobair]

Hi Guys,

I have one comment here,,

Although IGP is not involved on the distribution of Labels as this is the function of Label distribution protocols, but Still (Traffic Engineering) information needs special LSAs and TLVs through OSPF and ISIS routing protocols.

So the Tunnel Label of the traffic Engineering in MPLS which is the TOP label, leads us to say still OSPF and ISIS special TLVs and LSAs helps in carrying those information.

HTH

Mohamed

[marikakis]

Mohamed,

You are right that a certain property (distribution of link states in this case) described as an implicit disadvantage in a certain scenario, can also be described as an advantage in another scenario. Link state protocols are appropriate for the distribution of TE information. However, my understanding is that OSPF and IS-IS are used for distribution of topology information and associated link properties/constraints in the TE scenario. I think RSVP signals the path and is responsible for carrying the label in the TE case.

Kind Regards,

Maria

[Giuseppe Larosa]

Hello Maria, Mohamed,

>> I think RSVP signals the path and is responsible for carrying the label in the TE case.

This is absolutely correct, MPLS TE extensions to OSPF and IS-IS refer to all those changes needed to run a constrained shortest path first CSPF, the labels are assigned by RSVP TE.

It is important to note that the label space in the node is a unique pool and that LDP, BGP with labels and RSVP TE picks a label from this unique common space.

The label has to be unique that is label 30 if used by LDP cannot be used also by RSVP TE.

The constrained version takes in account the bandwith pools per direction on MPLS TE enabled links and the state of current reservation to calculate a path for the tunnel.

Hope to help

Giuseppe

[marikakis]

Giuseppe, thanks for taking the time to confirm the understanding of these concepts and for providing additional information.