Announcement

Collapse
No announcement yet.

TPPS: A New Linux Kernel I/O Scheduler

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

  • TPPS: A New Linux Kernel I/O Scheduler

    Phoronix: TPPS: A New Linux Kernel I/O Scheduler

    The Tiny Parallel Proportion Scheduler (TPPS) is a new I/O scheduler for Linux to appear on the kernel mailing list...

    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
    How would you go about benchmarking a scheduler anyway? Run 10 processes and see which one finishes them fastest?

    Comment


    • #3
      Yep, latency and throughput with many clients. Some of the dbench tests for example.

      Comment


      • #4
        Maybe along with other out-of-tree schedulers, such as BFQ (http://algo.ing.unimo.it/people/paol...ed/results.php)...

        Originally posted by Ancurio View Post
        How would you go about benchmarking a scheduler anyway? Run 10 processes and see which one finishes them fastest?

        Comment


        • #5
          Promising

          This sounds very promising!

          Comment


          • #6
            Not sure what they did with the ubuntu 12.04 scheduler.. its not that good compared to other versions judging by how files copy slower and all. It gets a very high io wait..
            Last edited by Caledar; 06 June 2013, 01:30 AM.

            Comment


            • #7
              Originally posted by VolandIT View Post
              Maybe along with other out-of-tree schedulers, such as BFQ (http://algo.ing.unimo.it/people/paol...ed/results.php)...
              I wonder how the two (TPPS and BFQ) compare in terms of performance for desktops?

              Comment


              • #8
                Originally posted by Ancurio View Post
                How would you go about benchmarking a scheduler anyway? Run 10 processes and see which one finishes them fastest?
                Depends on what matters most to you: Minimum/Maximum Latency, fastest completion of a specific app, fastest completion of a single app, fastest completion of all apps, CPU overheat, etc.

                Lowest latency, obviously, would be a pure priority scheme. FIFO has lowest CPU overhead. Highest total throughput would be Shotest Time First. Multilevel Queue and Round Robin both try to give acceptable throughput and acceptable latency at the cost of CPU overhead. And so on and so forth.

                In short, the optimal scheduler for your system is determined by your system requirements. A RTOS is going to use Priority. A general purpose OS is going to use some form of Round Robin or Multilevel Queue. And so on.

                Comment


                • #9
                  Ok, what happened to this scheduler? Was it abandoned?

                  Comment

                  Working...
                  X