Announcement

Collapse
No announcement yet.

A New BFS "Smoking" Scheduler For Linux 3.3

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • A New BFS "Smoking" Scheduler For Linux 3.3

    Phoronix: A New BFS "Smoking" Scheduler For Linux 3.3

    Con Kolivas announced this weekend the release of an updated BFS scheduler for the recently-released Linux 3.3 kernel. The new BFS scheduler is at version 0.420 and is codenamed "Smoking", with "a fairly large architectural change" since earlier versions of this out-of-tree kernel scheduler...

    http://www.phoronix.com/vr.php?view=MTA3Njg

  • #2
    Nice version number...one laid back mellow scheduler this version

    All kidding aside, this scheduler does work quite well IMO and here's to its continued improvement

    Comment


    • #3
      Originally posted by Paradox Uncreated View Post
      The regular kernel can do 0.33ms latency now. Where would this patch be applicable currently?
      My ass it can. BFS is still more responsive than CFS.

      Peace.
      War.

      Comment


      • #4
        He has a quad core, not dual. And the improvements can be felt but also measured. You can find plenty of results and graphs.

        So no. He is not wasting his time. For a lot of people, Linux wouldn't be the same without CK's work.

        Btw, writing "peace" at the bottom of your posts does not make you look cool. Unless you're Jamaican. Otherwise you're just a poser.

        Comment


        • #5
          Mike's benchmarks say it all

          https://lkml.org/lkml/2012/3/25/38

          I think those benchmarks really give a good indication of the current state of CFS vs BFS.

          CFS is actually more fair now, at the cost of worst case latency. It does have higher overall throughput.

          BFS is somewhat unfair (but still generally within a percent or two of being completely fair) and yet has much lower latency in some test (as much as 100% less latency). But the cost of achieving that low latency is hurting throughput and overall maximum number of threads.

          So, use what's best for your workload. I'd say CFS has made many improvements, and is really a generally good scheduler.

          Where as BFS is still the best scheduler if all you care about is low latency.

          EDIT: I also just read an attached academic paper, and it confirms that BFS does give very good fairness in worse case type scenarios that aren't predictable. BFS does well in workloads where the caches can't be guaranteed to work, e.g. cache misses (sounds like Desktop workloads to me). Where as CFS does well in predictable throughput based benchmarks, and it scales to lots of cores. BFS theoretically wouldn't scale too far since it has a global runqueue.

          http://research.cs.wisc.edu/adsl/Pub...n-thesis11.pdf
          Last edited by hechacker1; 03-25-2012, 03:23 PM.

          Comment


          • #6
            Originally posted by RealNC View Post
            My ass it can. BFS is still more responsive than CFS.


            War.


            Originally posted by RealNC View Post

            Btw, writing "peace" at the bottom of your posts does not make you look cool. Unless you're Jamaican. Otherwise you're just a poser.
            I'm just gonna say it, you are being a huge asshole for no reason. Calm down buddy.
            Last edited by bwat47; 03-25-2012, 03:17 PM.

            Comment


            • #7
              Originally posted by Paradox Uncreated View Post
              Why is it that some people go online, and write really lame-ass responses. "war". LOL. In my mail it said "fucking war". You removed the "fucking". Dude the whole thing is really lame.
              It is true, I also got the same text in my email.

              Originally posted by Paradox Uncreated View Post
              And your repulsive ignorance with "peace" etc, is inexpressibly stupid. Get educated, learn about some other cultures. And you will find it is a common greeting, and in groups far larger than jamaicans, rastafari, and hasish smoking.
              +1


              RealNC, do you feel cool if you write that? Get some peace, man.

              Comment


              • #8
                worst case latency

                It seems to me that some people are not placing enough value on worst-case latency for user interactivity and responsiveness. Most of the benchmarks I have seen here measure throughput and average latency. When I am using my computer, the worst feeling is if it "freezes" for a few seconds, even if that only happens once an hour, it really bothers me. I would be willing to give up a lot of throughput (or tolerate higher average latency) if it can eliminate such freezes. I don't know that BFS can accomplish that (or linux realtime, or something else), but whether another scheduler could eliminate such freezes or not, I don't think the benchmarks being discussed do a good job of measuring this issue.

                Comment


                • #9
                  Yo, peace mah man. Me an' mah homies are chillin' an' readin' thah forums. Yo, check it out man, people are writin' "peace" cuz it makes 'em look cool.

                  As for BFS, it's still better than CFS. I get XRUNS with JACK when running CFS that I do not get with BFS. So it's not a placebo, it's actually audible.

                  Word. (/me crosses hands over my chest and makes the peace sign)
                  Last edited by RealNC; 03-25-2012, 06:15 PM.

                  Comment


                  • #10
                    Originally posted by jwilliams View Post
                    It seems to me that some people are not placing enough value on worst-case latency for user interactivity and responsiveness. Most of the benchmarks I have seen here measure throughput and average latency. When I am using my computer, the worst feeling is if it "freezes" for a few seconds, even if that only happens once an hour, it really bothers me. I would be willing to give up a lot of throughput (or tolerate higher average latency) if it can eliminate such freezes. I don't know that BFS can accomplish that (or linux realtime, or something else), but whether another scheduler could eliminate such freezes or not, I don't think the benchmarks being discussed do a good job of measuring this issue.
                    I agree that worst-case responsiveness is probably more important for desktop use than average or best-case, but anything that's multiple seconds has NOTHING to do with a CPU scheduler. Maybe the IO scheduler.

                    Comment

                    Working...
                    X