[e2e] [Fwd: State of MPLS deployments today]

[Jennifer Rexford]

> It is not surprising that MPLS ends up targeting a narrow application,
> basically a form of traffic engineering.  Answering the three previous
> questions gets much simpler when the "edges" are the edges of a provider
> network, rather than the edges of the Internet, and when the timescale
> over which the decisions are made is expressed in days or months, rather
> than seconds or milliseconds: a provider can obtain information on
> network usage by statistical analysis of traffic over a long period; the
> edge points can be coordinated by a form of central planning; and the
> decisions are made by the very owner of the resource.

Nicely put.  Still, this kind of medium-to-long time scale TE doesn't
necessarily require the flexibility (and associated complexity) of
MPLS and the TE extensions to OSPF/IS-IS.  Tuning the configuration of
conventional IP routing protocols (like OSPF/IS-IS with "static" link
weights set at the network-management timescale) can often get you
very close to the "optimal" distribution of traffic achievable with
explicit routes.  For a recent survey of work on this approach, see

  http://www.research.att.com/~jrex/papers/ieeecomm02.pdf

from the October 2002 issue of IEEE Communication Magazine.  This is
not to say that MPLS can't "add value" but rather that it is not
strictly necessary for good traffic engineering.

-- Jen