Considering switch from Black Diamond 6808 to Catalyst 6500

Mikael Abrahamsson swmike at swm.pp.se
Wed Jun 15 15:25:52 UTC 2005


On Wed, 15 Jun 2005, Drew Weaver wrote:

> inexplicable/random lag that only seems to effect VoIP/Game Server type
> applications. We're just trying to find opinions to see if A) anyone is
> using these switches successfully, B) if anyone had problems with them
> and switched, or C) if anyone has ever heard of a switch that even
> though its 97% idle it still has performance issues :D

Unless you have brand spanking new linecards and the MSM3, all ICMP 
packets will be CPU routed on the BD (and all Extreme Networks i-chipset 
platforms).

We successfully use BD6808 with ICMP hardware routing blades and MSMs, 
plus the ip-subnet-lookup feature in EW7.3 which brings it up from a 
"route-cache is a /32 type platform" into giving it an additional network 
size (we use /16) route-cache with 128k entries. These changes enables it 
to being able to route todays internet.

If the CPU is 97% idle I don't really see that you should have this 
problem, the above two solves the issue where netTask (which handles 
forwarding tables being programmed into ASIC plus CPU forwarding of 
packets) is very high, which is not your case.

If you don't use it as a router, none of the above applies to you, the 
BD6800 is an excellent L2 switch that should never have any problems what 
so ever unless you flood it with multicast (a la slammer).

How do you measure the latency, that you say is 40ms? If you can measure 
it with anything else than ICMP then you might see other results.

I am behind several right now and we're running voip and gaming over them.


About changing, if you're planning to move to 6500 with Sup720-3B(XL) then 
you'll get a very capable platform, which is at least one generation newer 
in performance etc. The BD6800 wwith i-chipset blades was released in 
1999. Personally, I like both of them, they're good at different things.

-- 
Mikael Abrahamsson    email: swmike at swm.pp.se



More information about the NANOG mailing list