[gclist] Mars Rover info from comp.risks
Thu, 08 Jan 1998 14:50:19 -0600
>In fact, what you *probably* want here is a priority boost for a thread that
>has just completed a wait. When the consumer releases the buffer lock, the
>producer, which has been waiting, should get the CPU next.
What's wanted depends deeply on problem. If you are trying to get things
done on scheduele there are complex problems no one has really solved.
In the simpler case where you are simply trying to increase thruput
every task should have a priority for every resource inversly proportional
to its use of that resource. This means a cpu hog needs low priority
for the cpu while a disk hog needs low priority for the disk.
If I'm going to make massive use of a disk and another task will be
in and out quickly it makes sense to let them go first because they
will then free up resources that other tasks need.
The good thing about the simpler case is that there are no hard
succeed fail rules only statistical measures on which you can do
more or less well. The existence of special cases where any given
system will do badly is acceptable.
Charles Fiterman Geodesic Systems
414 North Orleans Suite 410 Phone 312 832 1221 x223
Chicago IL 60610-4418 FAX 312 832 1230
Object orientation without garbage collection is nonsense.