> This does not pertain to linux-kernel in any way:
Huh? Apparently you don't run gated very much...
> Ascend has announced
> the "GRF 400", an IP "switch" (huh? what? an IP switch?). It
> promises hardware assisted route table lookups and traffic across all
> interfaces at "wire speed". The specs mention that is can handle
> 150,000 routes.
> http://www/ascend.com/products/grf400/grf400index.html
Speaking of which, does Ascend have SNMP working on their boxes yet? Can
you still crash them by portscanning them?
> For those of us that are bandwidth-hungry....this is definitely a
> "woodening" article.
Provided you think that they can do it. I think that a more reasonable
approach is that presented at http://www.pluris.com
There's a whole bunch of people out there throwing vaporware promises
around about "solving" the routing table problem, in the hopes that enough
industry momentum behind them will give them enough time and money to
"solve" the problem. This is a fundamentally hard problem, born not of
Unix architecture (notice that it is IOS which is having the problem) nor
of insistence on using "stock" hardware (there's enough specialized
hardware in your average cisco7k to power the space shuttle.)
This is a _fundamentally_ hard problem, and no one, at least that I have
seen, has a solution. "hardware assisted route table lookups" may be
interesting, but I for one have not seen Patricia trees implemented in
hardware.
I'll believe one of these promises when I see it work. Until then, I'd
rather work on making sure that Linux copes as well as possible with these
challenges.
__
Todd Graham Lewis Linux! Core Engineering
Mindspring Enterprises tlewis@mindspring.com (800) 719 4664, x2804