[kwlug-disc] Stress-testing network switches

[unsolicited]

Paul Nijjar wrote, On 05/05/2010 6:06 PM:
> This is on topic, sort of. 
> 
> We have a bunch of older network switches. I worry that some might be
> flaky. I would like to stress-test them, preferably with FLOSS tools. 
> 
> My idea is to take a couple of laptops, put iPerf or netpipe on them,
> and blast traffic between the laptops on each port of the switches. 
> 
> Is there a better way? Are there better software tools I should be
> looking at?

I'll assume, given your message, that these are not smart switches 
wherein the switch itself can give you stats.

And I don't suppose you have any hardware test equipment to hand. 
Stuff that does more than connectivity / crossover / pair testing, not 
being cheap.

Real life tends to be the best stress test. Especially if you are 
sensitive to the possibility of a flaky switch, you'll be more 
inclined to switch ports faster should an issue develop. (You're ready 
for the shoe to drop, if it should.) It can also be more user 
irritating - less so if they're advised you're testing / there could 
be potential problems.

I suppose it depends upon whether you suspect the switch backbone, or 
particular ports. If you suspect the backbone, you would need to 
simultaneously nail multiple ports (and have corresponding 
destinations of the test data). Switch fabrics / backbones being 
significantly faster than any one port, usually.

And it probably depends upon to what use you would like to put the 
switches. e.g. If you're going to use a single switch in an one-off / 
by itself, or give it away, that's one thing, where you really want to 
be sure it's all happy before pawning off a potential problem to 
someone else.

OTOH, if you'd like to make use of them in your switch stack, then 
putting two on top of your stack, and using real life testing (being 
sensitive and ready to swap ports at the first sign of trouble), may 
not be unreasonable. If it turns out the whole switch is flaky, you 
have the second in place for testing anyways.

If the issue might be heat, then location may be the only restriction 
to use. (Ceiling / hot room, vs. under your desk as a temporary test 
switch.)

Sorry I don't have a more specific answer for you, but I suppose some 
of the answer depends upon your success, or good enough, criteria.

Also, be sensitive to the idea that some older switches don't really 
auto-negotiate all that well all the time - particularly GB adapters 
to MB ports. Depending upon the equipment attached to a particular 
port, a printer running at 10 Mbps half duplex, may be sufficient, for 
example.

By any chance does the switch manufacturer, like a hard drive 
manufacturer, have any test utilities you could take advantage of?