Subject: Re: bitching about ciscos (Was Re: Rolling in IPv6 code into -current)
To: Peter Galbavy <email@example.com>
From: Rob Healey <firstname.lastname@example.org>
Date: 04/28/1996 13:26:48
> I haven't yet had time to look at the propsals on the table for the next
> NANOG, but people tell me that it involves yet another round of magic
> restrictions to fix the problems *caused* by the massive presence of
> Cisco. "Apparently" (in quote, cause we all knew this, insteresting how
> others are only now finding out) their CPUs cannot cope with the route
> flap going on out there, so they are going to fix the problem by providing
> facilities to drop prefixes base on some inversely exponential metric to
> do with the prefix length.
> sigh. Cisco are proud to have 80% of the Internet infrastructure they say.
> Can they fix their intentionally under powered systems to actully allow
> the world to keep working ? No. sigh.
> I am glad we use NetBSD boxes for routers. I can even put up with the
> instabilities of gated.
I guess I would like rock solid proof that current mainstream router
vendors OTHER than Cisco could do any better. While I'm not a big fan
of Cisco they have solved problems that others have yet to solve and I'm
not willing to let the OTHER vendors off the hook until I see
new experiences to replace my old experiences with Bay, 3com, etc.
I've seen Bay, 3com and others fall flat on their face in large
company WAN/LAN's with only thousands of IP routes, let alone the
tens of thousands of routes the Internet has to deal with.
I WOULD be willing to bet that a UNIX based system would keep going
longer than the stand alone boxes since most UNIX systems now
out there have much more resources than the router boxes.
Routing table and update problems go beyond just Cisco 7000 series
not being able to handle more than tens of thousands of route updates.
There is also a MAJOR difference between being a regional network and
being part of the top level route structure, things get sticky when you
have to make sure tens of thousands of routes get to the right place
at the right time. i.e. this stuff is a bitch when you try to scale
in to monster stomping size! Ask MCI, Sprint, Alternet and others
how much "fun" running backbones can be...
Simply blaming Cisco for scaling problems won't help solve the problem.
And trying to convince thousands of corporations they should replace
Cisco's (or any other stand alone router) with UNIX boxes probably
won't go over too well. The end solution will be a new breed of
monster stomping stand alone router boxes that have even more resources
than a maxed out UNIX box and work in Gbit per second quantitys!