[e2e] EC++N

J. Noel Chiappa jnc at ginger.lcs.mit.edu
Wed Apr 3 08:53:12 PST 2002

    > From: Jon Crowcroft <Jon.Crowcroft at cl.cam.ac.uk>

    > the routing system should damp _decisions_ made on the basis of suc=h
    > signals- 5 mins might be a bit long - it wil ldepend on the actual rate
    > of change of number of flows on various paths....  

Ah, prior experience in a number of systems that did load-dependent routing
(the ARPANet, and the US long-distance network) has shown that it's very,
very, easy to get the damping wrong, and produce oscillatory routing
behaviour. (E.g. link A is loaded, and B is unloaded, so all the traffic
moves over to B, leaving B loaded, and A unloaded. Repeat.)

Given the problems we are already having with routing stabilization in the
current routing architecture (which is a large, distributed computation that
has to stabilize), I would suggest that adding the additional impetus of load
information at the wide-area level is probably not wise (and I'm not sure how
much use it would be at the local level).

I think a much better architecture is one that "nails up" a 'carrier' for a
traffic aggregate along the entire path and then normally doesn't move it;
new 'carriers' can be laid out and installed avoiding loaded elements. Such a
design would be immune to load-induced oscillation. (There are still some
second-order problems having to do with overall congestive collapse due to
non-efficient use of resources when the network is a whole is highly loaded,
but I'm going to gloss over them.)

Since the damping constants are very tricky to get right, and the correct
values often depend on the details of the actual network and load patterns
(see the literature - sorry, don't have citations available offhand), such
an architecture is really very preferable.


More information about the end2end-interest mailing list