[net.unix-wizards] automatic niceing by kernel

george (11/09/82)

berkeley 4.1 seems to automatically renice processes to 4
after they have run for a certain length of time (~ 12 minutes).
i thought it was the line in setpri() commented 'nice 4' but
that doesn't seem to have anything to do with cpu time usage.
it would nice if it was a constant somewhere but it 
doesn't seem to be that either. any help would be greatly appreciated.

chris.umcp-cs@Udel-Relay@sri-unix (11/19/82)

From:     Chris Torek <chris.umcp-cs@Udel-Relay>
Date:     15 Nov 82 01:46:31 EST  (Mon)
That's in "clock.c".  If CPU usage goes too high, and the process
isn't owned by UID 0, and the processes current nice is 0, it gets
put at four.  Interestingly, you can always "renice" it to one, and
it stays there.  (We have a modified renice here that lets anyone
renice his own processes to nonnegative values.)

rlb@Purdue@sri-unix (11/22/82)

From: Bob Brown <rlb@Purdue>
Date: 19 Nov 1982 00:27:04-EST
We felt that the automatic renicing of processes by the kernel was
slightly bogus.  All it did was clog up the proc table with alot
of pig processes (big troff, NA stuff, simulations) and just degrade
the system as much as if they all just ran to completion.  The code
in clock.c was the first thing we yanked out of 4.1BSD.  It especially
caused problems here because we serialize troffs...if one got reniced,
the queue of batched jobs behind it would get longer and longer and longer
and eventually we were faced with overnight turnaround on troffs.

Bob Brown
Purdue-CS