Announcement

Collapse
No announcement yet.

Should Ubuntu Use The BFQ I/O Scheduler?

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

  • Should Ubuntu Use The BFQ I/O Scheduler?

    Phoronix: Should Ubuntu Use The BFQ I/O Scheduler?

    The BFQ I/O scheduler is working out fairly well these days as shown in our benchmarks. The Budget Fair Queueing scheduler supports both throughput and low-latency modes while working particularly well for consumer-grade hardware. Should the Ubuntu desktop be using BFQ by default?..

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Is BFQ better than NOOP or Deadline on SSDs?

    Comment


    • #3
      Originally posted by uid313 View Post
      Is BFQ better than NOOP or Deadline on SSDs?
      what metric do you want to be better? throughput, latency, desktop responsiveness?

      Comment


      • #4
        Originally posted by boxie View Post

        what metric do you want to be better? throughput, latency, desktop responsiveness?
        I think in general most people want to see desktop responsiveness

        Comment


        • #5
          Originally posted by boxie View Post

          what metric do you want to be better? throughput, latency, desktop responsiveness?
          Hmm, good question. I had not considered that. I guess desktop responsiveness.

          Comment


          • #6
            Originally posted by boxie View Post

            what metric do you want to be better? throughput, latency, desktop responsiveness?
            Does porn load quicker?

            Comment


            • #7
              Looking at Michael's latest round of tests, BFQ ranges from dramatically improves startup times, to about the same to way worse in FlexibleIO or PostgreSQL compared to no scheduler at all. Until it can improve some areas without negatively impacting others, I don't see a reason to switch.

              boxie Isn't low latency the same as desktop responsiveness?

              Comment


              • #8
                Originally posted by bug77 View Post
                Looking at Michael's latest round of tests, BFQ ranges from dramatically improves startup times, to about the same to way worse in FlexibleIO or PostgreSQL compared to no scheduler at all. Until it can improve some areas without negatively impacting others, I don't see a reason to switch.

                boxie Isn't low latency the same as desktop responsiveness?
                I see the desktop responsiveness as a mix of latency vs throughput. somewhere in the middle where we sacrifice some throughput perf for other apps being able to function and not being blocked - but still getting high throughput for those tasks that need it (e.g. loading a level and having a webpage load from cache at the same time)

                Comment


                • #9
                  BFQ uses non-standard cgroupfs control files (for example
                  Code:
                  blkio.[B]bfq[/B].weight
                  instead of
                  Code:
                  blkio.weight
                  ). It breaks I/O resource management as configured in systemd. Before this is fixed in BFQ, it shouldn't be default.

                  Comment


                  • #10
                    I have been using bfq for a few months now on my laptop, and temporary switching back to noop and using the system made me notice the pauses/stuttering when the system is opening apps or copying data to a USB stick.
                    Going back to BFQ I don't notice the stuttering of desktop anymore.

                    So, for a desktop, BFQ sure does _feel_ a lot faster.

                    Comment

                    Working...
                    X