The MPLS WG Archive

[Shahram Davari <Shahram_Davari@pmc-sierra.com>]

Hi,


> Q1. Is the rule, "Use per-platform label space for
> inner label of tunneled packets" is true for ALL cases ?

No. If somehow you know from which interface an assigned label will arrive, we could use per-interface labels. The reason that I said use per-platform label is because MPLS is mainly used for TE, so the LSP could change dynamically and therefore the interface that the label arrives could change.
> 
> Q2. Is the following approach for allocation of label for
> inner label incorrect ?
> 
> Consider the following case:
> 1. R1 & R5 are two LSRs controlled by enterprise1
> 2. R2, R3 and R4 are three LSRs controlled by provider1
> 3. LSP Tunnel L1 originates at R2 and terminates at R4
>    a. Label between R4 & R3 is L1 (from interface label space)
>    b. Label between R3 and R2 is L1' (from interface label space)
>    c. Assume that the LSP has been advertised as link into ISIS/OSPF
> 	R1----------R2-------R3-------R4-------R5
> (LSP L1)		----L1'------L1----	
> 
> Now, if one has to create a LSP L2 between R1 and R5
> can the following scheme be used ? 
> 
> 4. LSP L2 originates from R1 and terminates at R5
>    a. Label between R5 and R4 is L2 (from interface label space)
>    b. Label between R4 and R2 is L2' (from interface label space)
> 	This Label L2' will be the second level label until R2.	
>    c. Label between R2 and R1 is L2" (from interface label space)
> 
> 	R1----------R2-------R3-------R4-------R5
> (LSP L1)		----L1'-------L1----	  
> (LSP L2)----L2"-----L1'|L2'-----L1|L2'---L2---

Yes. This works as long as the LSP tunnel L1 is static or at least terminates at the same R4 interface.

> 
> Q3. You also pointed out that following: "Also note that any label
> that has been drawn from the per-platform label space should not 
> be used any more by the per-interface label spaces."
> 
> The LSR MIB (lsr-mib-08) states that the 
> mplsInterfaceParticipationType for a specific MPLS Interface can
> either be perPlatform(0) or perInterface(1). Does one have to take
> out label drawn from the per-platform label space, from the label
> spaces of interface that are not participating in the perPlatform
> label space ?

I don't understand what you mean by "interface that are not participating in the perPlatform label space". If you mean that you are 100% sure that no per-platform label could arrive at a particular interface, then you probably don't have to exclude them from that per-interface label space. But a router has millions of labels, which I doubt will be fully used. It probably is safer to exclude the per-platform labels from all per-interface label spaces.


Regards,
-Shahram