Re: [PWE3] [mpls] ELI as a reserved label

[Curtis Villamizar <[e-mail]>]

In message <[e-mail]>
Shane Amante writes:
>  
> Curtis,
>  
> On Jul 30, 2010, at 14:02 GMT+02:00, Curtis Villamizar wrote:
> > In message <[e-mail]>
> > Shane Amante writes:
> >> On Jul 30, 2010, at 11:06 GMT+02:00, Curtis Villamizar wrote:
>  
> [--snip--]
>  
> >>> 3.  Providing a means to carry indication of large flow as described
> >>>     in draft-yong-pwe3-enhance-ecmp-lfat-01.txt and
> >>>     draft-yong-pwe3-lfc-fat-pw-01.txt without changing the use of TC
> >>>     in a forwarding LSP label.
> >> 
> >> I'm unclear how a single, reserved label would be a means of signaling
> >> a large vs. small flow.  How can a single label represent two states
> >> (large vs. small flow)?
> > 
> > That's the slightly ugly part.  If ELI carries a TTL=0 then it can't
> > be used to forward and among those that discussed this (in one of the
> > "hallway sessions" near the registration desk) it was not seen as too
> > horribly ugly if TC meaned something else in the ELI label stack entry
> > only.
> > 
> > Lucy has a different way to do multipath and this is an enabler.  If
> > your LSRs don't use it, then it does no harm as long as the hashed
> > entropy value is there too.
>  
> Understood.
>  
> I would note that in draft-kompella-mpls-entropy-label-01 it is NOT envisioned that there will be a "standalone ELI" in all cases.  Specifically, in the case where application labels are in use, (e.g.: 2547 VPN's, 6PE, etc.), the application label itself is intended to serve the purpose of an ELI, indicating that the label that immediately follows the application label is an entropy-label, i.e.:
>  
> +----------------+
> |  Tunnel Label  |
> +----------------+
> |   App. Label   |
> +----------------+
> | Entropy Label  |
> +----------------+
>  
> In these cases, there wouldn't be a reserved label used for a ELI in the label stack.  Thus, I'm not sure you can count on the LFC-bit *only* being set in entropy-labels that are immediately preceded by a reserved ELI label.
>  
> So, with respect to LFC, IMO it's:
> a)  orthogonal to the discussion of whether or not a reserved label for the ELI is absolutely _required_ or just nice-to-have; and,
> b)  the MPLS WG needs to weigh in with a definitive answer on whether changing the semantics of the MPLS TC is appropriate to accommodate end-users who desire LFC.
>  
> -shane


This doesn't work in the following cases:

  1.  P2MP (multiple egress, therefore multiple ELI values).

  2.  Label stack depth is limited in order to accommodate MPLS-TP
      (midpoints must know to look past the ELI to include the entropy
      label in the hash).

Since Lucy's large flow work needed TC, using TTL=0, and reserving one
TC bit for large flow seemed to do no harm.

I spoke to Stewart and Kireeti about this at IETF.  Perhaps either or
both of them could weigh in on this topic on the list.

Curtis