MPLS question

thegreattriscuit

First, this HAS been covered... a bit in the thread here, but it's quite old and I'm not sure how y'all feel about thread necromancy here, so I'm starting a new one.

This stems from a particular paragraph (and assosciated figure) in the ISCW Exam Certification Guide (Cisco Press) that simply defies my interpretation and understanding of the whole point of MPLS.

as follows:

The LSP is essentially a tunnel between source and destination for a particular FEC. Zooming out, it would look more like a tunnel with many forks in the road, because it is feasible for two FECs to share a label at one point, then diverge at another.

I've also attached the figure.

My question is simply: HOW!?!? if the whole concept and purpose behind MPLS is that any non edge LSR will forward based ONLY on the label of an incomming packet, then if at any point two FECs use the same inbound label, wouldn't they, from that point on, be effectively the same FEC and have no opportunity to diverge until the labels were popped?

So, are these statements simply wrong? or am I completely loosing my mind here?

Any insight would be appreciated, before my head explodes, thanks


actually, it has occured to me that there's an easy way to visualize what I'm talking about:

(NOTE: periods are spaces, it won't let me leave whitespace, so I have to use periods...)

Based upon MY understanding, you could have this:

FECA-->--->--->\
........................>FECC>--->--->--->Edge LSR
FECB-->--->--->/

but NEVER this:

FECA-->-->-->\................... /-->FECA-->
......................>FECC>-->-->
FECB-->-->-->/ ...................\-->FECB-->

which is pretty much the opposite of what the book is saying, as far as I can tell.

kalebksp

I can't think of a situation where that could occur either. The closest thing would be if another label was pushed on top of the original labels of two different FECs. The two FECs would have the same top label for some period of time, but when the top label was popped off they could again have differing top labels.

cisco_trooper

Labels are locally significant if I remember correctly. Those two subnets have the same next hop along the MPLS path so they are using the same label.

networker050184

cisco_trooper wrote: »

Labels are locally significant if I remember correctly. Those two subnets have the same next hop along the MPLS path so they are using the same label.

A router doesn't use the same label for all routes via the same next hop. In regular unicast MPLS forwarding a unique label is assigned to each prefix even if they have the same next hop as far as I know.

This part also confused me during my studies for MPLS. I don't know enough about MPLS to know all the situations that could occur like this. Maybe with something like MPLS TE, but I'm not sure.

thegreattriscuit

A router doesn't use the same label for all routes via the same next hop. In regular unicast MPLS forwarding a unique label is assigned to each prefix even if they have the same next hop as far as I know./QUOTE]

Well, such was my understanding aswell, and try as I might, i have not been able to replicate that behavior in dynamips/GNS3, and the more I think about the way label distribution has been described to me, the less and less that paragraph seems likely, so I guess I'll just write it off as a slipup on the authors part and hope for the best come test time

Thanks for the help guys, I just really needed a sanity check there before I take this exam next week :P