No announcement yet.

Linux 4.12 I/O Scheduler Benchmarks: BFQ, Kyber, Etc

  • Filter
  • Time
  • Show
Clear All
new posts

  • Linux 4.12 I/O Scheduler Benchmarks: BFQ, Kyber, Etc

    Phoronix: Linux 4.12 I/O Scheduler Benchmarks: BFQ, Kyber, Etc

    Among the many new features for Linux 4.12 are two new I/O schedulers in mainline: the long-standing BFQ (Budget Fair Queueing) and Kyber, a new I/O scheduler developed at Facebook. Here are some initial benchmarks of these I/O schedulers on the Linux Git code as of this past week.

  • #2
    Sorry Michael, but would it be possible to see these benchmarks on a HDD?


    • #3
      So I'm assuming (and that's a dangerous thing to do) that the "noop" scheduler of the first set of benchmarks is the same as the "none" scheduler in the second set of benchmarks?

      Additionally, it would be nice to see CFQ in the second set of benchmarks since that's the default scheduler that the vast majority of systems are actually running.


      • #4
        so noop is actually better?


        • #5
          How do you benchmark the pros of BFQ.. i.e responsiveness for desktop users in I/O intensive situations?


          • #6
            Don't you need to measure interaction and out-of-context latency to actually see the effects of these schedulers?

            Surely noop/none is and always will be ideal for an isolated operation with no interaction...

            The reason to use a scheduler is to prevent something from stealing all of the resources and making the system miss events (e.g. input), right?


            • #7
              I am not a performance expert, but I will share this thought. It seems like "artificial" performance tests, that test a specific niche area, are only useful when compared to themselves, as in previous versions. This would allow you to see progress over time. It might be interesting to label "real world" tests so that they stand out. With scheduling, I can see some tests pointing to an overall victor, who, in reality, is not the best for real workloads.


              • #8
                Interesting. Hopefully somebody does similar benchmarks on the old spinning-rust hard drives that everyone actually uses when large amounts of data storage are needed.


                • #9
                  wow bfq performs horrible , maybe paolo can fix it for 4.13
                  i like the responsiveness of bfq so i would like to see it performing well


                  • #10
                    All of these benchmarks measure throughput...

                    I wonder how it would be possible to measure latency/responsiveness. Run some IO-intensive task in background, and run some pre-recorded GUI click-thorugh using mouse/keyboard emulation, and point camera at the screen and measure when things pop-up? (Or record video using the videocard, or take screenshots).

                    Alternatively, run some 3D game + some CPU/HDD intensive background task, and measure min-FPS?

                    Write some synthetic latency/responsiveness benchmark which measures time it takes to read 1st byte from file or get a CPU slice?

                    As per old saying- whatever is not measured, is lost... Until we have a decent way to measure latency/responsiveness, it will continue to be mostly ignored as a concern...