Announcement

Collapse
No announcement yet.

OpenZFS 2.1-rc5 Released With Linux 5.12 Support, Many Bug Fixes

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

  • OpenZFS 2.1-rc5 Released With Linux 5.12 Support, Many Bug Fixes

    Phoronix: OpenZFS 2.1-rc5 Released With Linux 5.12 Support, Many Bug Fixes

    Two weeks have passed since OpenZFS 2.1-rc4 while today a fifth release candidate was issued for this open-source ZFS file-system implementation for Linux and FreeBSD systems...

    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

    I was thinking you could do a filesystem benchmark (zfs vs ext4) with some things such as timing these kind of tests:

    7z ultra compress this big file: http://mirror.siena.edu/almalinux/8....x86_64-dvd.iso

    tar bzip2 the linux kernel source files

    generate the same file dir trees on both filesystems and then do a find with bit pattern search






    Comment


    • #3
      One more filesytem performance test:
      Chia farm of a certain size copied to both filesystems(same farm data), k=35 at least, then time and run "chia plots check"

      Comment


      • #4
        Originally posted by onlyLinuxLuvUBack View Post
        I was thinking you could do a filesystem benchmark (zfs vs ext4) with some things such as timing these kind of tests:

        7z ultra compress this big file: http://mirror.siena.edu/almalinux/8....x86_64-dvd.iso

        tar bzip2 the linux kernel source files

        generate the same file dir trees on both filesystems and then do a find with bit pattern search

        ext4 would win that kind of test though, the kinds of things ZFS excels at are DB reads with it's arc. ZFS probably will never win in a write test. (because it does more work writing checksums, it's kind of a meaningless test.. if ext4 calculated the checksums then it would make sense..)

        90% of filesystem benchmarks are designed to defeat the cache.. and that's unfair to ZFS because it's cache is an integral part and is not a standard LRU cache like most other filesystems. Oracle themselves have had this problem with professional benchmarks.

        In real world ZFS preforms very well but synthetic.. no. You have to keep your workload in the arc to see good numbers and usually the arc does a good job doing so.

        Arc After Dark: Kind of long but a must watch to understand ZFS performance.
        Bryan Cantrill on "ARC: A Self-Tuning, Low Overhead Replacement Cache" by Nimrod Megiddo and Dharmendra Modha ( https://www.usenix.org/legacy/event/fast03/te...
        Last edited by k1e0x; 12 May 2021, 03:56 PM.

        Comment

        Working...
        X