hanszorn2000:

> A typo that Helen Borrie copied into her book as well then, unfortunately.
> But assuming this is a typo, how can we explain the fact that when I
> increase the value of this parameter from 100 ms to 400 ms, it gets
> much worse? I then expect my active threads will be turned to low
> priority later, giving them more time to finish (which should be well
> possible within that timeframe). When a time-taking query is running
> at the same time, the performance of the small jobs drops
> dramatically. As if the priority was low already...
>

Well, this is not a typo, but part of a document, taken in initial
letter out of context. Full text:

# The wait time, in milli-seconds (ms), before the priority of:
# - an active thread is reduced to 'Low', or
# - an inactive thread is increased to 'High'
#
# Note: The default value was chosen based on experiments on Intel
# PIII/P4 processors. It should be increased for using in the computer
# with lower speed processors.
#
# Type: integer
#
#PrioritySwitchDelay = 100

Everything is correct - when thread becomes active it's priority is
reduced to Low after given timeout, when it becomes inactive - increased
to High after same timeout. Making this parameter too big can easily
make the things worse.