Announcement

Collapse
No announcement yet.

Should Ubuntu Use The BFQ I/O Scheduler?

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

  • #41
    Originally posted by starshipeleven View Post

    yes
    .
    .
    you sir, are stirring shit :P

    Comment


    • #42
      The BFQ dev seemed to be implying that the changes in 5.2 make BFQ preferable for even NVMe drives and did away with its major read+write problem areas. Excited to see what your new benchmarks say.

      Comment


      • #43
        Originally posted by Britoid View Post

        Does porn load quicker?
        No, but you can wank while you are defragmenting you filesystem ;-)

        * This post is sponsored by the Low Effort Internet Jokes Gang!

        Comment


        • #44
          Originally posted by uid313 View Post

          Do people still use those these days?
          They are excellent data storage.

          Comment


          • #45
            Originally posted by tildearrow View Post

            Yes. I still do, and many people around here do too.

            Why can't you understand you simply can't drop support for hardware that is still being actively released only because a new technology came out?
            Yeah, I know SSD tech has been out in the market for like 9 years, but they still make HDDs.
            I am not saying support should be dropped for HDDs or BFQ should be removed, but maybe most people these days have SSDs and that maybe BFQ isn't such a good default choice for most people.

            But maybe it is possible for the system to determine if the storage device is a HDD or a SSD and set the appropriate scheduler.

            Comment


            • #46
              Not for server because "BFQ doesn't perform well for PostgreSQL database server workloads."

              Comment


              • #47
                Originally posted by Space Heater View Post
                By the way bfq does inherit ionice priorities from the task's CPU niceness if no IO priority class is set, this has been the default since it was first introduced.
                See: https://git.kernel.org/pub/scm/linux...osched.c#n4895
                Cool, then the documentation is just out of date, and I am running into another issue. I have had pretty good success with just adding an explicit ionice level to my icecc daemon though.

                Comment


                • #48
                  Why don't you try cgroup? very easy. There's a script to make it even easier*. It's only cgroups v2 (unified) for now, but that's the future anyway

                  Also, cgroups have much higher granularity than ionice/nice. Values are 0-10000 by default.

                  * https://github.com/Gatak/cgexec

                  Comment


                  • #49
                    Just tried it on my Fedora 30/ssd laptop and I don't know if that's just bias but feels quite a bit more responsive when having running (io-heavy) stuff in the background. Will keep it on for a while and see.

                    Comment


                    • #50
                      Originally posted by boxie View Post

                      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)
                      Neah, desktop is all about low-latency. If you're loading a game level while browsing, you won't notice if the loading takes a few more seconds.
                      Although, truth be told, I'm not so sure why I need the low-latency either. My first PC ran MS-DOS and only later it got Windows 3.1/3.11. To me the current DEs are pretty nifty as they are.

                      Comment

                      Working...
                      X