On Wed, 15 Mar 2000, Alex Belits wrote:
>On Wed, 15 Mar 2000, Horst von Brand wrote:
> Last time I have checked, scheduler handles priority of processes
>without any "time hog detector" in userspace, yet nice(1) is used all the
>time. It can argued that daemon can help to set policies for running
>processes, but the possibility to at least set policy on the program
>startup by some wrapper should improve situation a lot compared to what it
>is now.
I think that's a good approach: use something much like Rik van Riel's
patch to kill processes in an OOM situation, but allow a user-space daemon
to alter the calculation he uses to determine which process to kill. If
we're OOM the daemon can't be run, so the daemon should not try to handle
OOM situations directly.
Here is an example (poor, but it demonstrates the idea): by default, when
OOM kill the process with highest x=memory/sqrt(CPU time). But each
process might have some priority value which is added to x. This priority
may be modified by root via a userspace daemon or a tool like nice.
The daemon might not be necessary; it would be better to set priority when
the task is loaded.
The only difference between this approach and Rik's patch is that it keeps
(some) policy in userspace.
Dave
David Whysong dwhysong@physics.ucsb.edu
Astrophysics graduate student University of California, Santa Barbara
My public PGP keys are on my web page - http://www.physics.ucsb.edu/~dwhysong
DSS PGP Key 0x903F5BD6 : FE78 91FE 4508 106F 7C88 1706 B792 6995 903F 5BD6
D-H PGP key 0x5DAB0F91 : BC33 0F36 FCCD E72C 441F 663A 72ED 7FB7 5DAB 0F91
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Mar 23 2000 - 21:00:20 EST