Re: Tree based scheduling

Rauli Ruohonen (raulir@fishy.pp.sci.fi)
Fri, 23 Jan 1998 09:23:53 +0200


In linux-kernel, Rik van Riel wrote:
>Please take a look at VSTa. It's had a tree scheduler
>right from the start. It's a very nice and simple scheduler,
>but response times aren't yet very good (haven't checked the
>latest version though).

You forgot to give a pointer, so I did a bit of searching and found this one:
http://www.chat.net/~jeske/VSTa/ (in case somebody else is interested)

[SCHED_LATENCY]
>It would be quite good. Except for one problem: every
>user will want to run each of his/her processes with
>this priority to get better response times :-)
>This might nullify the effect :-(

Hmpfh, this needs more thought.. In any case, this priority would be useful
even if it were root-only, because it's kind of "safe semi-realtime"
priority. Maybe root could do more fine-tuning, i.e. set the time-slice
limits for individual SCHED_LATENCY processes (in case you are on your
personal machine and want to give mpg123 some priority but not so much that
it'll halt everything else).

>> SCHED_BATCH could also be good, it would also be just under the
>
>Excellent idea... But what do you do when a SCHED_BATCH
>process blocks because of a pagefault?
>You wouldn't want to bring in another SCHED_BATCH process
>then, would you?

Something else (not SCHED_BATCH process) could be run, if possible. If not..
:/ This needs more thought, too..

>Maybe a normal (ie: as found in other unices) swap daemon
>(ie: true swapping) solve this problem far better?

What does the swap daemon do? I haven't used other unices much..

-- 
The only difference between a car salesman and a computer salesman is
that the car salesman knows he's lying.