Announcement

Collapse
No announcement yet.

The Performance Of Ubuntu Software Running On Windows 10 With The New Linux Subsystem

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

  • #21
    Originally posted by carewolf View Post
    And Ntfs [...] is still a decade more modern and an order of magnitude faster than the best Apple has to offer.
    Didn't Apple plan to include zfs eventually? What happened to that?

    Comment


    • #22
      Originally posted by dragorth View Post
      Okay, Microsoft has had an LVM since XP. It just isn't used by default on an install. Why? So users could install Linux.
      Originally posted by carewolf View Post
      Windows had a fully functional LVM layer long before Linux did. One problem though was that with Win2k and XP you didn't get it with the home editions. I haven't checked recently.
      I'm not talking desktop OS, I'm talking Server. Microsoft does not have a proper LVM even on their Server OS's. What LVM are you talking about?? Because if it's Windows Disk Manager.... LOL... just LOL. That's like a speak-n-spell My First LVM. It's so bad, Veritas offers their VxFS and Veritas LVM for Windows servers and charges $thousands for it. Windows Disk Manager does not do ANY of the core functions of a real LVM. Can you grow your C:/ live on the fly no downtime? No. Can you migrate an LV from one PV to another? No. Can you do all kinds of creative striping, mirroring, and spanning? Nope, you can only do a 2-member mirror OR a 2-member stripe (but not both) and creating either of them is a one-way function. So please, what is this Windows LVM you speak of, because this bad joke called Windows Disk Manager can't possibly be it.

      Comment


      • #23
        Originally posted by Kushan View Post

        I have no doubt that there are certain workloads that favour one filesystem over another (and it doesn't help that NTFS is ancient by today's standards), however even worst case it's not 100x slower.
        It is not the filesystem itself which is slower, as evidenced by ntfs-3g not having these problems. It is the Windows FS layer.

        Comment


        • #24
          Thanks for the benchmarks - I am genuinely surprised at the speed this preview has! I hope this work is open sourced (NFI if it is or plans to be). If it is I can see some clever people doing a very simple OpenGL "pass thru" driver so we can all get our Linux desktop and user space on windows

          Comment


          • #25
            Originally posted by chithanh View Post
            It is not the filesystem itself which is slower, as evidenced by ntfs-3g not having these problems. It is the Windows FS layer.
            To be a fair comparison we would need to have an additional windows FS benchmark (without any Linux layer)

            Comment


            • #26
              Originally posted by torsionbar28 View Post



              I'm not talking desktop OS, I'm talking Server. Microsoft does not have a proper LVM even on their Server OS's. What LVM are you talking about?? Because if it's Windows Disk Manager.... LOL... just LOL. That's like a speak-n-spell My First LVM. It's so bad, Veritas offers their VxFS and Veritas LVM for Windows servers and charges $thousands for it. Windows Disk Manager does not do ANY of the core functions of a real LVM. Can you grow your C:/ live on the fly no downtime? No. Can you migrate an LV from one PV to another? No. Can you do all kinds of creative striping, mirroring, and spanning? Nope, you can only do a 2-member mirror OR a 2-member stripe (but not both) and creating either of them is a one-way function. So please, what is this Windows LVM you speak of, because this bad joke called Windows Disk Manager can't possibly be it.

              Its called Dynamic disks.

              I didn't say it was better, just that they had it. And, yes you can grow your live c:\ drive with no down time.

              Keep in mind that Windows Disk Manager is just that, a program that manages disks. It isn't the actual LVM, merely and interface.

              Just like Windows Explorer is an interface to the NTFS file system, but doesn't make use of all features. NTFS can have mind bogglingly huge filenames, but Explorer is limited to 255 characters for legacy reasons. (Bash can create these huge filenames that Explorer can't handle, so this is a heads up.)

              Comment


              • #27
                While I personally have zero interest in this Lin-on-Win weirdness - which seems perversely inverted, I'm a bit alarmed by what struck me as some shocking Ubuntu regressions... with 16.04 often managing only around 50% of 14.04's computational performance... and on one benchmark barely achieving 33%

                It looks the problem could be caused by the (newer) Linux beneath 16.04. (?) Anyone know what's going on???
                Last edited by Dick Palmer; 15 July 2016, 02:22 AM. Reason: oopsie!

                Comment


                • #28
                  Perhaps the perf issues have something to do with Intel CPU drivers?

                  Comment


                  • #29
                    I think the explaination here is that CPU bound applications don't make system calls. So the Linux subsystem in those tests is actually mostly not involved. The really interesting tests are probably those using the filesystem, and comparing them to native windows test would be a good way to actually test the brand new layer. Interesting workloads would also be those of parallel applications heavily using locks (semaphores etc).Since those need system calls, they would stress the new subsystem. I see also that in many tests windows shows a very high variance compared to bare linux (it is that +-x right?). Is it possible to have 90/95/99th percentiles figures with those benchmarks?

                    But we should also considering that hardcore performance is not the goal here, nobody will run high performance stuff on top of this; I believe it is mostly intended to be able to test if something works, not if it is fast (actually, just grep on windows is enough to justify the effort )

                    Comment


                    • #30
                      Originally posted by Dick Palmer View Post
                      While I personally have zero interest in this Win-on-Lin weirdness which seems perversely inverted, I'm a bit alarmed by what struck me as some shocking Ubuntu regressions... with 16.04 often managing only around 50% of 14.04's computational performance... and on one benchmark barely achieving 33%

                      Since the Lin-on-Win userland is 14.04 but didn't suffer the regressions it seems the problem is caused by the (newer) Linux beneath 16.04. (?) Anyone know what's going on???

                      Usually it's the powermanagement of some kind that fsck's CPU performance over. Tweaks in P-state, CPUFreq, Turbo boosting etc have the nasty tendency to never get perfected. Some machine works on some kernels and then gets fubared by some other kernel. My recommendation is that you disable every thing that has to do with powersaving etc. That will atleast get you to a state of consistent performance. (Don't know about turbo boost etc, idiotic feature btw.)

                      Comment

                      Working...
                      X