Scheduler in Linux kernel

[rohitgupta14]

Which scheduler is currently in use in fedora.. or any recent linux kernel?
when will CFS be a default scheduler for linux [http://kerneltrap.org/node/8059 ]

[leigh123linux]

Its CFS

* Fri Jul 06 2007 Dave Jones <davej@redhat.com>
- 2.6.22-rc7-git6

* Fri Jul 06 2007 Dave Jones <davej@redhat.com>
- Remove another common wakeup (cursor blinking).

* Fri Jul 06 2007 Dave Jones <davej@redhat.com>
- Update CFS to v19.

* Fri Jul 06 2007 Jeremy Katz <katzj@redhat.com>
- Add minimal patch from markmc for ICH9 support in e1000 while the new
  driver works itself out upstream

[rohitgupta14]

Which scheduler was in use before CFS?
CSF was released on this Date: Fri, 13 Apr 2007 22:21:00 +0200

[tekel]

I am pretty sure CFS has been around longer, it was just put in the mainline on 13 Apr

[ibbo]

Basically

Preemptive scheduling is the normal Linux way to schedule. Its a time slice based schedule where each process is given (N) amount of time to perform its task. If it does not complete it must wait untill the schedule comes back around to continue its run (like when something hangs for a second, or a window jiters when moving it). Of course depending upon your system you may find different algorithms in play. For instance real time system would use perhaps a rate monotonic algorithm to do scheduling.

I have yet to look closely at CFS but the buzz is on so it cannot be worse than whats been before. Perhaps it will even lead to the death of the real time linux Kernel (who's goal was also to provide efficient scheduling).

Ibbo

[tekel]

When you say just dolling out time slices you are skipping the large and complicated matter of who get slices and how many they get. Linux does have a very good priotorization scheme, O(1) and the such in the current scheduler, You are really only getting foundational changes with CFS not real performance changes. It's been said CFS will preform just as good and the vanilla scheduler for now, but it should give the kernel devs some room to tweak things and hopefully get much better performance in the future.

[ibbo]

Oh my yes I did, but then I do not have the time the energy or the patience to delve into the intricacies of such algorithms or how the Kernel actually slices up its processes (old and new) and delegates a slice for each be it on priority for kernel space or user space apps or even new process's introduced.

But in my own defence I did say Basically!.

Of course a quick google can easily produce page after page of meaningless text and equations (unless you know whats what) that can give put you in geek heaven (or hell if you decide to write your own scheduling algo).

Ibbo

[rohitgupta14]

http://kerneltrap.org/?q=node/341

This is the scheduler that was in use before CFS, known as O(1) scheduler...