Linked by Thom Holwerda on Sat 11th Aug 2007 20:46 UTC, submitted by anonymous
Thread beginning with comment 262730
To read all comments associated with this story, please click here.
To read all comments associated with this story, please click here.
RE[2]: 0.10 scheduler comment
by flanque on Sun 12th Aug 2007 04:03
in reply to "RE: 0.10 scheduler comment"
RE: 0.10 scheduler comment
by butters on Sun 12th Aug 2007 06:33
in reply to "0.10 scheduler comment"
This is a pretty good algorithm, all things considered. Remember, Linux won't support SMP for another four years. Traversing the task list 2-4 times isn't great, but O(N) is perfectly acceptable for dozens of tasks. Some commercial UNIX schedulers still scale linearly.
They also still use the "C = C/2 + P" function to give sleepers a bounded priority boost over hogs. Maverick sysadmins can tune the denominator these days, but the theory is identical. Yes, the algorithm is non-deterministic fuzzy math, but it's simple and it works.






Member since:
2005-07-06
I particularly enjoyed this comment, on kernel/sched.c:
* 'schedule()' is the scheduler function. This is GOOD CODE! There
* probably won't be any reason to change this, as it should work well
* in all circumstances (ie gives IO-bound processes good response etc).