[c-nsp] Extreme vs. Cisco

[Drew Weaver]

On 3/29/06, Jon Smith <netguyster at gmail.com> wrote:
> I was wondering if anyone has any good comparison of cisco switches
> vs. extreme? Also any horror stories about extreme. I have used
> Cisco's products and like them very much, but need to argue to
> management that we don't need to go the Extreme way.
>
> Any thoughts would be appreciate it.
>
> --Jon
>

The rhyme is "Extreme is trouble free until layer three". 

	It all depends on your environment, the number of entries in
your forwarding information database, and the version of software you're
running, Also the version of modules, and the platform (summit, 6808..
et cetera) plays a big role however, the point is many of the factors
wouldn't matter if it were a well developed piece of equipment. To the
environment end, I'd say that if you were an organization such as a
hospital, an office building, or another sort of company which can
control the type of packets flowing in and out of it; then it could
possibly suit your needs, however it is not suitable for any sort of
hosting environment where you must 24/7 be able to take whatever anyone
throws at you.

	Secondly, the forwarding database issue; the forwarding database
on the extremes is limited because it directly taxes the CPU and if you
are dealing with a large number of prefixes (destinations and or source
addresses) you will run into issues unless you aggregate or use
proprietary extreme commands to "limit the number of prefixes the CPU
will deal with" this again is due to poor design on the software and
hardware end of things. About a year ago we made a conscious decision to
stop using Extreme gear and migrate to Cisco Catalyst 6500s. We have
since started mixing 6500s and 4000s (Sup720s and SupVs) in different
places, and we couldn't be happier.

Some of the other random issues we had with the black diamonds are.

--pretty much everything is handled by the CPU so when people are doing
"performance testing" against your network it always appears to be
slower than it is (ICMP, etc) [We all know that pings are not a sign of
performance, but to many hosting customers, this is what they go on.]

--The amount of PPS the black diamond (at least the 6808) can handle
before performance degradation at Layer3 is somewhat low.

--VOIP, Gaming, and any sort of application which employs UDP showed
significant loss compared to similar tests using Catalysts.

	That is my summary of our experience with Black Diamonds in
Layer3, as I said we still employ the exact same units in Layer2 and
they work beautifully as forwarders.

FWIW,

-Drew