Re: Missing pieces in draft-team-tewg-restore-hierarchy-00.txt

[Sudheer Dharanikota <sudheer@nayna.com>]

Jim:

One comment inline.

Jim Boyle wrote:

> Sudheer - you raise an interesting question...
>
> On Thu, 16 Aug 2001, Sudheer Dharanikota wrote:
>
> > Hi all:
> >
> > I have following comments on the design team draft presented at
> > IETF. I already discussed some of these comments with
> > Wai.
>
> <snip>
>
> > 3. Although you say only horizontal hierarchy is considered, we propose
> > atleaset two layer (Client and Server layer) capabilities should be
> > considered.
> > Example, if the server layer is optical or SONET/SDH then it itself can
> > possess some restoration capabilities, which need to be considered by
> > the client layer technologies. This helps in choosing the protection and
> >
> > restoration mechanisms wisely.
> >
>
> It was felt in the design team that carriers are not quite at the point of
> deploying GMPLS into their optical cores, not quite yet.  And even if they
> had GMPLS cores, ready to be tapped into by routers from *within* their
> company - most felt a bit uneasy about this prospect.
>
> Are there folks deploying GMPLS Optical Cores that are worried that there
> will be a time real soon when their routers won't be able to signal for
> bandwidth on demand?  Or that there might be some crucial piece of
> communication their Optical Core might be able to confer to their Routers
> that might change their behavior (that couldn't be conferred in a more
> ready manner - such as by configuration of the router) ??
>
> The area that did seem more pressing was how to support edge to edge
> signaling across a data network, which might have OSPF areas.  There
> was concern about scalability, and a desire to have folks look into
> approaches which would allow edge to edge signaling in a manner which
> is more scalable than possible today.  The recurring context for such
> a need was VPNs (mpls presumably) over IP/MPLS networks.  The need to
> signal edge to edge might very well be questioned, however some folks
> seem to think that if a VPN product offers an SLA, this is somehow
> necessary.
>
> Anyone else think that the "vertical" hierarchy is pressing (or a
> non-issue at this time)?
>

The issue I have raised is not just for the vertical hierarchy in my opinion.
GMPLS is mainly used for making transmission equipment more intelligent.
The intelligence is in selecting a path through these equipment (includes both

path computation and path setup). People are known to pre-provision different
transmission topologies to recover quickly from failures. Path computation
needs accurate representation (or accurate abstraction :-) ) of the
transmission
layer topology and its capability.   This in my opinion is not captured in the
draft.
I am still talking about a single layer here.

Regards,

sudheer

>
> Anyone have interesting operational perspectives which clarify the
> "horizontal" requirements?
>
> regards,
>
> Jim