Re: Performance Tests Cisco 2514 and Pentium 166/400

From: Nick Bastin (nbastin@mil3.com)
Date: Tue Feb 29 2000 - 17:36:22 EST


>We do it quite often :-)
>
>We have customers running 2 very loaded T3s in PCs so you are just plain
>wrong about that. (see the 110K pps discussion from the other day). I
>suppose that the DS3 wan cards you have heard of are not up to snuff. Our

I'm obviously hoping that's it. This was also not 2.2.x, so there must be
more improvement there than I thought. I will be doing some tests in the
near future with linux boxen and IP routing over high-bandwidth low-latency
links, so at that point I'll probably be soliciting optimization tips from
this list.

>dual HSSI card can run fully loaded quite easily, with bandwidth management
>(although I wont say you can have 1000 rules). A 500Mhz box seems to be
>able to handle it quite nicely.

What I've found is that to a certain extent, as long as your bus and cpu
can push the bandwidth, you can have as many rules as you want. Copying
the packet from the card to ram is expensive...running that copy through
the CPU a few more times is cheap, considering how much more bandwidth you
have between the CPU and the memory bus, as compared to the PCI bus. As
you point out, you can go overkill with this, but you can have a *lot* of
rules before that becomes your bottleneck.

>You seem a bit biased towards the 7206...even more so than Cisco itself :-)

Naw, not really. I've never personally touched a 7206 (many a 7513,
though), although I do believe that it can do when the spec sheet says its'
capable of. My only bias towards actual routers as opposed to PC's in
general is that I've only had Cisco hardware crash on me once, and never
die (of hardware failure), whereas I can't say that about my PC firewalls
and routers. Not that I've ever had a linux firewall crash the OS on me,
but I do frequently have the hardware just up and die. Lets just say that
I get more of that 'warm fuzzy feeling' from my routers than my linux
boxes....something about no moving parts. (Except fans of course, but at
least the fan in your cisco isn't close enough to it's own cable to the
point where you bump the case and it will eat its' cable and disconnect
itself and then proceed to overheat the CPU and kill the machine...Murphy
lives with me, I swear.. :-))

--
Nick Bastin
Software Developer
OPNET Technologies

- To unsubscribe from this list: send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.rutgers.edu



This archive was generated by hypermail 2b29 : Tue Mar 07 2000 - 21:00:26 EST