[10785] in North American Network Operators' Group
Re: Internet Backbone Index
daemon@ATHENA.MIT.EDU (Sean M. Doran)
Mon Jul 14 13:02:16 1997
To: Stephen Balbach <stephen@clark.net>
Cc: "Dorian R. Kim" <dorian@blackrose.org>, nanog@merit.edu
From: smd@clock.org (Sean M. Doran)
Date: 14 Jul 1997 12:28:52 -0400
In-Reply-To: Stephen Balbach's message of "Sun, 13 Jul 1997 20:15:48 -0400 (EDT)"
Stephen Balbach <stephen@clark.net> writes:
> Routes the first packets and switches the rest based on
> "flows".
How is this conceptually different from cisco
fastswitching wherein the slow path is used for the first
packet towards a destination and all subsequent packets
use a cache-based fast path?
In both cases you have a slow cache-set-up path and a
finite cache.
There are significant disadvantages to using a cache in
core routers (look up Dennis Ferguson's comments in your
local NANOG archive, nobody I can think of can improve on
them).
The only principal advantage to cache-based routing is
when you haven't got the power to process all packets
through the same path without drops or mishandling of
things like options and filters, but you do have the power
to do some sort of hashed cache lookup. With luck, this
will not be the case in future routers.
Sean.
