Skip to contentSkip to conversation

New scheduler in 2.4. series?

DISCUSSION Andreas Tscharner 4 participants 9 messages Feb 4, 2002 – Feb 6, 2002

lore
Share & tools
lore
#0 lore raw

Hello World,
Hello Marcelo,

Is there any chance, that the new scheduler will be in the 2.4 series soon?

Regards
Andreas

Signature · 5 lines
Andreas Tscharner                                     [email protected]
-------------------------------------------------------------------------
"Programming today is a race between software engineers striving to build 
bigger and better idiot-proof programs, and the Universe trying to produce
bigger and better idiots. So far, the Universe is winning." -- Rich Cook 
#1 lore raw

Seeing the new scheduler in 2.4 would be nice! However, soon wouldn't be nice, because there are a lot of 3rd party kernel modules that try to make calls against the old scheduler that don't seem to work. My personal case in point is Bestcrypt ( www.jetico.sci.fi ). Just my thoughts on the matter.

--Brandon

-=-=-=-=-=-=-=-Previous Message(s)-=-=-=-=-=-=-=-

Quoted 13 lines

Hello World,
Hello Marcelo,

Is there any chance, that the new scheduler will be in the 2.4 series soon?

Regards
Andreas

Signature · 5 lines
Andreas Tscharner                                     [email protected]
-------------------------------------------------------------------------
"Programming today is a race between software engineers striving to build 
bigger and better idiot-proof programs, and the Universe trying to produce
bigger and better idiots. So far, the Universe is winning." -- Rich Cook 
#2 lore raw
Quoted 3 lines · Brandon Low

Seeing the new scheduler in 2.4 would be nice! However, soon wouldn't be nice, because there are a lot of 3rd party kernel modules that try to make calls against the old scheduler that don't seem to work. My personal case in point is Bestcrypt ( www.jetico.sci.fi ). Just my thoughts on the matter.

If you're speaking of Ingo's O1-scheduler I won't vote for inclusion now. I still have some trouble with high priority nice levels (renice -20). For some seconds the system gets totally unresponsive for user requests while switching between those processes. The last one I've tried was J2.

*Kristian

:... [snd.science] ...:
::
:: http://www.korseby.net
:: http://gsmp.sf.net
:.........................:: ~/$ [email protected] :

#3 lore raw
Quoted 3 lines · Kristian

[...] I still have some trouble with high priority nice levels (renice -20). For some seconds the system gets totally unresponsive for user requests while switching between those processes. [...]

if you are using nice -20 tasks then they might take CPU time away from lower priority tasks. This is why bigger negative nice levels should only be used sparingly. (and this is why it can only be done as root.)

Quoted 1 line

The last one I've tried was J2.

that's a pretty old patch, the current one is -K2.

Ingo

#4 lore raw
Quoted 3 lines · Ingo Molnar

if you are using nice -20 tasks then they might take CPU time away from lower priority tasks. This is why bigger negative nice levels should only be used sparingly. (and this is why it can only be done as root.)

Of course. But with the good old scheduler I've never had any problems. The system was very slow but continuously responsive. (low frequency timeslices between process rotations) With your -J2 patch it's getting really unusable. Maybe I should give -K2 a try.

*Kristian

:... [snd.science] ...:
::
:: http://www.korseby.net
:: http://gsmp.sf.net
:.........................:: ~/$ [email protected] :

#5 lore raw
Quoted 2 lines · Kristian

Of course. But with the good old scheduler I've never had any problems. The system was very slow but continuously responsive. [...]

true, the old scheduler did not do much with the nice level. In the new scheduler there is some real difference between nice levels - and this is a feature. I'd suggest for you to try something like nice -5. What kind of application are you renicing?

Ingo

#6 lore raw
Quoted 4 lines · Ingo Molnar

true, the old scheduler did not do much with the nice level. In the new scheduler there is some real difference between nice levels - and this is a feature. I'd suggest for you to try something like nice -5. What kind of application are you renicing?

Just multiple 'nice -n -20 tar -cvjf` in a row for "high priority backup when my DTLA drives makes wierd noises again". I'm just compiling -K2. (BTW: -J2 behaves really bad while moving windows during a kernel compilation.)

*Kristian

:... [snd.science] ...:
::
:: http://www.korseby.net
:: http://gsmp.sf.net
:.........................:: ~/$ [email protected] :

#8 lore raw
Quoted 3 lines · Kristian

-K2 behaves much better as -J2 did. The system gets continuously responsive again. But with nicelevel -5 at the beginning of each 'tar` the systems stalls for 2 or 3 seconds.

yes, it takes 2-3 seconds for the system to notice that the 'tar' process started from your interactive shell is in fact a 'CPU hog'. The system was honoring root's request for CPU time.

this should not happen if you start it at nice -2, which should still give 'tar' enough of an advantage.

Ingo