nice

bruce at graffiti.UUCP bruce at graffiti.UUCP
Wed Mar 12 15:31:06 AEST 1986


Approximately how much 'control' does a process have over the processor
when nice'd as high as it can go?  Does anybody have hard numbers?

You've got to get a big job done as quickly as possible.  The system 
is already loaded fairly heavily.  If you just run your job as normal,
it will run very slowly and make everyone else's jobs run slowly also.
If you could just tell unix to only do the one high priority job
and don't even switch between it and another task, you cut down 
a huge amount of overhead (especially if the system would otherwise
start to thrash).  Your high priority job would halt everyone
else's jobs completely for a short period of time
while the system cranks out your task in a much more efficient manner,
then after everyone is finished cursing you, they get on with their
tasks with a normal response time.  After all, if it takes vi a couple
of seconds for every response, you may as well not even be using it.
You could use your time to do something else.

It appears that nice is not as effective as it could be (at least on
my system).  It appears that the time required to run all processes
concurrently is much, much greater than the combined times of running
the big jobs separately from the little ones which run concurrently.

Sorry if this is basic material.  (Is this called job control, which
I've been seeing in the Subject field of articles I haven't been reading?)



More information about the Comp.unix mailing list