Some things learned:
There are two schedulers, the IO scheduler and process scheduler.
That's not news, but the fact that you can't Google for one without
being confused by the other, is. Especially when they just call it 'the
scheduler' and don't get explicit.
The UT/ET/et al problem isn't induced by IO; it's not a disk-request
being serviced for 3s or some network activity. Cycling through all
available choices (without resorting to patches) has no effect on the
behaviour.
When it comes to _process_schedulers_ there's not a lot of detail,
other than a world of submitted patches and a lot of talk starting in
April 2004, and continuing today. But I see no word of finalization and
choices like for IO. :(
For those of you with a need for IO scheduling, all the details
you'd want (and more, trust me) are located in the kernel tree under the
'block' directory. Yes, the kernel tree was sitting there all this
time. And now that I'm sure I'm not the only one who feels stupid for
forgetting it :) I'm off to research _process_ schedulers there.
More as I discover it.
--
------------------------------------------------------------------------
Brian FahrlÃnder Christian, Conservative, and Technomad
Evansville, IN http://www.fahrlander.net
ICQ 5119262
AIM: WheelDweller
------------------------------------------------------------------------
Attachment:
signature.asc
Description: This is a digitally signed message part
