[c-nsp] Extreme vs. Cisco

[Richard A Steenbergen]

On Fri, Mar 31, 2006 at 09:24:34AM -0500, Drew Weaver wrote:
> 
> 	Another good example of this is during SQL Slammer's
> introduction we were using Black Diamond 6808s; and I remember all of
> our Black Diamonds raising to 100% CPU/RAM utilization simultaneously
> due to un-patched/unmanaged boxes being connected to them and sending
> thousands of connections to 'unroutable IP addresses' 299.x.x.x, etc.
> and the black diamond was happily trying to pass them along
> unsuccessfully forever. I've not seen this behavior with the Catalyst.

If you don't understand the performance characteristics of random 
destination worm-scan traffic on a non-prepopulated forwarding table you 
are doomed to failure, on any platform. There are plenty of Cisco 
platforms that have this same issue, as well as many many other vendors 
(not to mention plenty of hacks to optimize fib/cam entries with or 
without complete prepopulation in order to reduce the effects of non-fast 
path initial lookups). Throwing out a blanket "Cisco doesn't have this 
issue" answer is very very very wrong.

Though in all fairness, Cisco did learn this particular on their L3 switch 
platforms first. If more people with common sense and demanding traffic 
loads had been willing to run L3 on the Extremes they probably would have 
had this brought to their attention sooner, but thats another issue. :)

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)