> The biggest challenge will be finding a test system with enough power to
> handle the production load.

I do most of my LISTSERV for unix testing on a 433MHz system, precisely
because I want to see performance problems before customers run into them.
We cannot test or even anticipate every possible real-life configuration
issue, but this helps. We do of course have much faster development systems,
but they are used mainly to test scenarios where a lot of horsepower can be
assumed, and for benchmarking. For the purpose of testing your LISTSERV
startup time with 1.8e vs. 14.4, you can use just about any system. Copy the
data, time with 1.8e, time with 14.4. My guess is that your startup times
will go down to well under 10 minutes with 14.4, if the only thing you
change is the software version. Having done that, you can then look into
tuning and, perhaps, minor hardware upgrades or reconfigurations. If you
only see a minor improvement with 14.4, it means there is some kind of
contention at the file-system level.

  Eric