Testing procedures for new network implementation?

Wed Aug 11 22:24:32 UTC 2004

On Wed, Aug 11, 2004 at 10:20:44AM -0400, Wayne Chow wrote:
> We are in the process of planning for the upgrade of a 3Com network.
> The new infrastructure will be comprised of 3Com 4950s with XRN and a dozen
> stacks of 4400s.
> What are the best practices of testing the new implementation?  Any advices
> or suggestions would be greatly appreciated.
> The only software that we have is the 3Com Network Director.

Ignoring the lame trolling which has already beset this thread, the 4400 is a 
solid, reasonably priced and fast edge switch [1]. 3Com could do some things 
better but then so can lots of manufacturers. They're certainly a world apart
from the old Superstack II stuff. Unfortunately they do little to market
some of the little know features of the platform. 

You want at least v3 firmware although v4 is out now and we've pretty much
deployed it with no problems.

There's not really a great deal to testing a L2 infrastructure. Analyse
the failure modes fully - power down devices to see what breaks, unplug optics 
and replug to check links come back up cleanly. Test and tune your spanning 
tree if you use it. Test multicast if you're using it. Create some loops
and see if storm control works. 

Usual site-specific stuff you'd probably have done during procurement -
test with your common desktop hardware, test with other devices, etc. 

We don't really use 3NS - it doesn't scale at all to our network - but we
do give it to some of the remote site 2nd-line people where they have a few
handfuls of switches in nonstandard topologies to look after.

Will.

[1] with ~400 24 and 48-port devices deployed on our network I feel I
can speak with a degree of authority. For our edge, Vendor 3 won a 
competitive OJEC tender vs others including Vendor C.