The MPLS WG Archive

[Tony Przygienda <prz@siara.com>]

Curtis Villamizar wrote:

> In message <Pine.GSO.4.21.0004271459030.27622-100000@petra.ee.surrey.ac.uk>, Ll
> oyd Wood writes:
> > On Thu, 27 Apr 2000, Curtis Villamizar wrote:
> >
> > > URL please.
> >
> > http://www.cs.umd.edu/users/georgeap/
> >
> > L.
>
> Thanks.  I have seen this.
>
> My only comment is that this is very good work but QoS routing != TE
> when TE is being applied to large aggregates of traffic in IP
> backbones (or any other backbone where large aggregates of traffic are
> handled by a relatively small number of extremely persisitent tunnels,
> whether tunnels provided by LSPs, ATM VPs, or lambdas).
>
> Curtis

The measurement of  trends for different low-pass filtering techniques
on the accurary of information you get on average in your nodes and % blocking of calls or
% of packet drops will be holding for TE as well {assuming we're talking here about
dynamic LSP establishment or OMP'ish kind of load-balancing solutions}.
Available bandwidth in TE is nothing else but QoS in one metric _in terms of
flooding_! Observe again, I'm not talking about computation algorithms you
apply to such information. This part you cannot deduct from this thesis, I agree
and will be specific to the engineering solution you pursuit ...

    thxn

    -- tony