Page 2 of 4 FirstFirst 1234 LastLast
Results 11 to 20 of 32

Thread: Debian GNU/kFreeBSD Benchmarks With Its New Kernel

  1. #11
    Join Date
    Oct 2007
    Location
    Under the bridge
    Posts
    2,128

    Default

    Quote Originally Posted by hdas View Post
    OMG, apache and sqlite are extremely important benchmarks - every desktop (including netbook and mobile phone) user keeps running them all the time. If your machine can't do more than a million requests or inserts per second, you are doom3d and have kregressed. </sarcasm>
    Win!

    (obligatory rant about stupid 10 char limit and 1' edit window)

  2. #12
    Join Date
    Oct 2010
    Posts
    10

    Default

    UFS defers fsync() or ?
    I don't see how the difference could be explained otherwise...

    ...and after 5min googling:

    This less important data loss case is the one which most BSD's, including
    FreeBSD and DragonFly, use for UFS write()+fsync(). Under UFS a
    fsync() does not issue a media flush, it simply issues the I/O
    and leaves the data sitting the drive cache.

    source: http://leaf.dragonflybsd.org/mailarc.../msg00005.html


    Would be nice if phoronix made this quick research and put that as comments

    Cause "blah is 12% faster than bleh", thanks, i can read the graphic, you can spare the extra keystrokes if you're going to say nothing

    Conclusion: ext4 is faster, UFS has a broken implementation.

    Finally, SQLite by default fsync() each transaction which is slow, but applications can just disable that, Firefox probably does it in many places.

  3. #13

    Default

    Quote Originally Posted by FireBurn View Post
    Rolls eyes

    Look some pretty kgraphs that tell you knothing knew or kinteresting

    PTS is not a substitute for a good article

    Graphs are nothing without analysis
    Stop this, please! Think about Gnome users for the gmoment.

  4. #14
    Join Date
    Dec 2007
    Location
    Edinburgh, Scotland
    Posts
    574

    Default

    Quote Originally Posted by kraftman View Post
    Stop this, please! Think about Gnome users for the gmoment.
    Stop being such a gNewSense

    Oh wait... that one's taken already

  5. #15
    Join Date
    Oct 2010
    Posts
    1

    Default

    Quote Originally Posted by hdas View Post
    OMG, apache and sqlite are extremely important benchmarks - every desktop (including netbook and mobile phone) user keeps running them all the time. If your machine can't do more than a million requests or inserts per second, you are doom3d and have kregressed. </sarcasm>

    Thats odd, normally those are the tests I would care about, when comparing a server OS. ray tracing? cdraw? who cares again?

    mysql performance, apache, postgres, memcache, php.

    Those are benchmark comparisons that would actully matter.

    side note, people still use KDE? really??!?

  6. #16

    Default

    UFS may perform many times better than EXT4 with some Linux kernel releases in SQLite, but with PostMark the EXT4 file-system definitely carries the advantage.
    man mount -> nobarrier

  7. #17
    Join Date
    Oct 2009
    Location
    Netherlands
    Posts
    41

    Default

    Quote Originally Posted by nitedog View Post
    Thats odd, normally those are the tests I would care about, when comparing a server OS. ray tracing? cdraw? who cares again?

    mysql performance, apache, postgres, memcache, php.

    Those are benchmark comparisons that would actully matter.
    The differences in the tests you refer to are big enough to warrant further investigation. Both kernels could benefit from the performance lead in the other kernel, because it can apparently be handled more efficiently.

    side note, people still use KDE? really??!?
    lol, your first post and immediately you start trolling

  8. #18

    Default

    Quote Originally Posted by nitedog View Post
    Thats odd, normally those are the tests I would care about, when comparing a server OS.
    You would if done properly. At least apache isn't interesting for you, I don't know about the second one. I bet if you use Ext2 or Ext3 and play with mount options you'll get much better results.

    side note, people still use KDE? really??!?
    Where did you come from?

  9. #19
    Join Date
    Oct 2008
    Posts
    3,076

    Default

    A good sqllite test would be nice, as mentioned earlier it's used in a lot of different software, like Firefox. Just testing the number of inserts you can do in 1 second is not a useful benchmark, though, and all you have to do to prove this is to look at the results. If the user experience was really that much worse, no one would have that FS feature enabled. (barriers, i think?) A useful benchmark would be to create a realistic workload. Maybe even trace the work Firefox is doing. Run a few insers and updates, along with several queries. Nothing just does straight inserts like that, though - if a program did want to, it would place all of them inside a single transaction, and still wouldn't show the performance problems this test has.

  10. #20
    Join Date
    Oct 2009
    Posts
    2,081

    Default

    And.... how many sqlite transactions do you actually DO in real life. Sure, ff uses it. That's nice. How many transactions does it do with it? 4? (intentionally left without context)

Tags for this Thread

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •