Slashdot Mirror


New Performance Mailing List for FreeBSD

An anonymous reader writes "FreeBSD's fast as is, but it appears that folks over there are going the extra mile to make sure that it continues to be the top dog according to this recent announcement. With 5.1 promising to have native threads and 5.1 only a few months out, it is really good to see performance being taken seriously."

23 comments

  1. I thought FreeBSD was already winning there. by mnmn · · Score: 3, Interesting


    FreeBSDs performance seemed to me to be quite close to say Linux 2.5 running on XFS. I couldnt tell which was better but I think both were quite close to the hardwares ability. I havent tried running many threaded apps, at least not under pressure..

    I hope they dont take focus away from simplicity and robustness here. Multithreading is tricky, look at all the problems Linux had.. As a customer, I wouldnt want multithreading at all if theres any risk to the robust architecture.

    --
    "Give orange me give eat orange me eat orange give me eat orange give me you." -Nim Chimpsky
    1. Re:I thought FreeBSD was already winning there. by Anonymous Coward · · Score: 5, Informative

      Doing threads right is tricky and is why it's taken over two years to get done right (more correct than Slowaris' threading too). It's close to finished and coming down the home streatch of development I think. If you want to know why Linux's threading is an ugly and bastardized hack, read this paper on KSE.

    2. Re:I thought FreeBSD was already winning there. by cpeterso · · Score: 1


      Where are the performance benchmarks demonstrating FreeBSD (4 or 5) dominance? I haven't seen FreeBSD used in any big time benchmarks..??

    3. Re:I thought FreeBSD was already winning there. by Anonymous Coward · · Score: 0

      Of course, with 2.6 coming out, linux will also have a much cleaner threading model with NPTL.

  2. This just in... by Znonymous+Coward · · Score: 3, Funny

    From the office of Iraqi Information Minister Mohammed Saeed al-Sahhaf :

    BSD isn't dead.

    (Unless you live in a cave, you will get this joke)

    --

    Karma: The shiznight, mostly because I am the Drizzle.

    1. Re:This just in... by jo42 · · Score: 2, Funny
      > FreeBSD is dying

      Who let the Iraqi Information Minister out of Baghdad?

    2. Re:This just in... by Anonymous Coward · · Score: 0

      Unless you happen to be the loser who posted this, you'll know who the the joke is.

  3. Experience by chrysalis · · Score: 3, Interesting

    I just love BSD.

    I used to go with Linux for everything a while back, but once you discover BSD, the ports system, IPF andPF, and the way everything is properly packaged, you just fall in love with BSD.

    But when it comes to performance, from an user point of view, Linux still looks better.

    My home workstation has both FreeBSD 5-current and Gentoo Linux. I lately installed a 2.5.x kernel (the current running one is 2.5.67-mm1), and I must say that I was really impressed.

    Not sure how latest Linux kernels perform in a server environment, but on a workstation, everything is very responsive. Even when there is a lot of I/O (local disk-to-disk backups) or CPU activity (compilation), KDE is always smooth. Under load, Windows can take some time to refresh their content, but as soon as something is typed on the keyboard, or clicked with the mouse, the effect is immediate.

    While FreeBSD is rather fast (it looks like there was a big speedup regarding disk I/O between FreeBSD 4.x and 5.x), I now find it slow when I switch from Linux to FreeBSD.

    Slow is maybe not the right word, but there is a lot of "lag". When the system is busy, there's a slight delay between an action with the mouse or the keyboard and its effect. This is especially noticable with Konqueror. And when there is a compilation going on, XMMS oftens stops playing properly (there are cutoffs in the sound) .

    I also tried Mnogosearch (MySQL 4 + cache mode) on both Linux and FreeBSD 5-current. Searches are as fast on both OS. But during indexation, when the cached daemon flushes its buffers, FreeBSD nearly freezes. I mean that even logging in through SSH becomes very long. On Linux, although the hard disk seems to turns a lot, the system keeps being responsive for other tasks.

    --
    {{.sig}}
    1. Re:Experience by Anonymous Coward · · Score: 1, Informative

      This is a problem with the current non-blocking implementation of libc_r. The native threading support (either 1:1/libthr or M:N/KSE) will solve this. libthr supposedly isn't production quality yet, but it is committed and many have noted that this addresses most of that threaded performance gap you speak of.

    2. Re:Experience by Maudus · · Score: 2, Interesting

      The comparison might be interesting once again with first 5-stable FreeBSD release. 5-current is impressively fast, but still it is not as fast as it can be. As a developer release it has a lot of debugging features turned on, many parts of code has not been optimized yet etc. -- I just managed to recovere accidentally deleted file (read: a day of work). Thanks god to cat and grep and, of course, FreeBSD.

  4. mysql ??? PostgreSQL! by Anonymous Coward · · Score: 0

    Why bother with mysql my friend when PostgreSQL is there.
    Performance arguments do not work, see the latest benchmarks.
    Mysql got whipped.

  5. Performance numbers from Soekris list by Anonymous Coward · · Score: 0

    Take a look at FreeBSD's performance on the popular soekris hardware... > >> FreeBSD 4.7 39.6 Mbps / 28.5 Mbps > >> FreeBSD 5.0 30.4 Mbps / 22.3 Mbps > >> OpenBSD 3.2 32.4 Mbps / 24.4 Mbps > >> Linux 2.4 41.6 Mbps / 32.6 Mbps > > > > Was this with polling enabled? Interesting numbers. The ipfw overhead is > > higher than I'd have thought. > > OK, I finally got around to do some testing with polling enabled: > > FreeBSD 4.7+DEVICE_POLLING,HZ=2000 52.1 Mbps / 39.4 Mbps -SU

  6. I like using FreeBSD by Gizzmonic · · Score: 1

    My own and others!

    --
    (-1, Raw and Uncut is the only way to read)
  7. freebsd usability by bwhalen · · Score: 0, Flamebait

    If they could only make patching easier, so everytime a bind, sendmail, or other hack is discovered, one doesnt have to;
    cvsup the source tree
    rebuild the system
    rebuild the kernel
    boot single user
    install new kernel
    install new system components
    mergemaster
    reboot again.

    This is a royal f***in pita, up2date looks sinfully easy compared to this.

    --
    Where do you want to be, What are you doing to get there.
    1. Re:freebsd usability by Anonymous Coward · · Score: 0

      If you read the patch, you can figure out what it changes and then rebuild only that part of the tree that needs to be changed.

    2. Re:freebsd usability by Anonymous Coward · · Score: 0

      Exactly. This person spends more time trolling slashdot than actually RTFM.

      I for one am happy that I don't have to deal with dependency hell!

    3. Re:freebsd usability by Anonymous Coward · · Score: 0

      Uhm...

      If you read the security advisories you get the commands to upgrade the affected portion of the system in a neat, cut 'n paste format. What could be easier?

      For example, from a sendmail advisory:

      Download the patch and the detached PGP signature from the following
      locations, and verify the signature using your PGP utility.

      # fetch ftp://ftp.freebsd.org/pub/FreeBSD/CERT/patches/SA- 01:57/sendmail.patch
      # fetch ftp://ftp.freebsd.org/pub/FreeBSD/CERT/patches/SA- 01:57/sendmail.patch.asc

      Execute the following commands as root:

      # cd /usr/src
      # patch -p /path/to/patch
      # cd /usr/src/lib/libsmutil
      # make depend && make all
      # cd /usr/src/usr.sbin/sendmail
      # make depend && make all install

      No rebooting needed. How exacly do you want this to be more easy?

    4. Re:freebsd usability by pooh666 · · Score: 1

      WHY did you guys not mod up this excellent answer???