Page 1 of 2 12 LastLast
Results 1 to 25 of 32
  1. #1
    Join Date
    Feb 2003
    Location
    Panorama City, CA
    Posts
    2,581

    FreeBSD 6.1 performance issues

    Ive been playing around with FreeBSD and i do like it's very lightweight considering the alternatives.

    On a dual AMD MP 2200+ and 2GB ram Raid5 10K SCSI drives, Its not the top of the line of all servers... but its still a good server.

    When compiling software (apache,php,mysql,etc etc) it seems very sluggish... So i found a benchmark here on the forum and gave it a go, heres the results from that.

    /dev/da0s1f 17741062 1542666 14779112 9% /usr

    Start Benchmark Run: Sun Nov 19 07:52:35 PST 2006
    7:52AM up 1:06, 1 user, load averages: 0.00, 0.00, 0.00

    End Benchmark Run: Sun Nov 19 08:03:06 PST 2006
    8:03AM up 1:16, 1 user, load averages: 13.19, 5.62, 2.69


    INDEX VALUES
    TEST BASELINE RESULT INDEX

    Dhrystone 2 using register variables 376783.7 8960948.2 237.8
    Double-Precision Whetstone 83.1 1125.6 135.5
    Execl Throughput 188.3 988.8 52.5
    File Copy 1024 bufsize 2000 maxblocks 2672.0 25070.0 93.8
    File Copy 256 bufsize 500 maxblocks 1077.0 9511.0 88.3
    File Read 4096 bufsize 8000 maxblocks 15382.0 158168.0 102.8
    Pipe Throughput 111814.6 1256372.8 112.4
    Pipe-based Context Switching 15448.6 56268.0 36.4
    Process Creation 569.3 2913.1 51.2
    Shell Scripts (8 concurrent) 44.8 281.9 62.9
    System Call Overhead 114433.5 1166551.9 101.9
    =========
    FINAL SCORE 86.1
    FreeBSD 1.somesite.net 6.1-RELEASE FreeBSD 6.1-RELEASE #0: Sun May 7 04:42:56 UTC 2006 root@opus.cse.buffalo.edu:/usr/obj/usr/src/sys/SMP i386
    So after compiling all the software, i formated the box and installed Centos 3.8 ran the benchmark and much to my surprise the score was in the high 120's.

    Can someone shed some sunlight on this?
    Remote Hands and Your Local Tech for the Los Angeles area.

    (310) 573-8050 - LinkedIn

  2. #2
    Join Date
    Mar 2003
    Location
    California USA
    Posts
    13,681
    Jeremy I have seen exactly the same. I didnt feel like bugging with it so we just installed centos for the client
    Steven Ciaburri | Industry's Best Server Management - Rack911.com
    Software Auditing - 400+ Vulnerabilities Found - Quote @ https://www.RACK911Labs.com
    Fully Managed Dedicated Servers (Las Vegas, New York City, & Amsterdam) (AS62710)
    FreeBSD & Linux Server Management, Security Auditing, Server Optimization, PCI Compliance

  3. #3
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    Quote Originally Posted by Steven
    Jeremy I have seen exactly the same. I didnt feel like bugging with it so we just installed centos for the client
    Yup, I think I was the client in this case . I can confirm that I did indeed experience this...
    MediaLayer, LLC - www.medialayer.com Learn how we can make your website load faster, translating to better conversion rates for your business!
    The pioneers of optimized web hosting, featuring LiteSpeed Web Server & SSD Storage - Celebrating 10 Years in Business

  4. #4
    Join Date
    Mar 2003
    Location
    California USA
    Posts
    13,681
    Side note:

    We tried on multiple servers for example:

    Dual p3 1gb ram ide drive got around 113 on centos, while on freebsd got less then 80

    Besides the benchmarks, the servers didnt feel as fast when doing pretty trivial jobs, untaring etc.
    Steven Ciaburri | Industry's Best Server Management - Rack911.com
    Software Auditing - 400+ Vulnerabilities Found - Quote @ https://www.RACK911Labs.com
    Fully Managed Dedicated Servers (Las Vegas, New York City, & Amsterdam) (AS62710)
    FreeBSD & Linux Server Management, Security Auditing, Server Optimization, PCI Compliance

  5. #5
    Join Date
    Feb 2003
    Location
    Panorama City, CA
    Posts
    2,581
    I so agree on the untaring part.. I thought it was the raid array, but on centos there was no problem at all.

    Seems *BSD has had problems with multi proc/core, not sure where to go from this now.
    Remote Hands and Your Local Tech for the Los Angeles area.

    (310) 573-8050 - LinkedIn

  6. #6
    Join Date
    Apr 2001
    Location
    Pittsburgh, PA
    Posts
    1,306
    Quote Originally Posted by Jeremy
    I so agree on the untaring part.. I thought it was the raid array, but on centos there was no problem at all.

    Seems *BSD has had problems with multi proc/core, not sure where to go from this now.
    Go to 6.2 - this is exactly the kind of thing that's being worked on in 6.x.

    Kevin

  7. #7
    Join Date
    Jun 2003
    Location
    Ukiah, California
    Posts
    394
    Further - what scsi card are you running? I have had horrible (IO) performance with some adaptec U160 cards. Switching to LSI cards fixed many of the problems.
    Jacob Turner -- Ringnebula Systems
    Managed IT solutions for small business

  8. #8
    Join Date
    Feb 2003
    Location
    Panorama City, CA
    Posts
    2,581
    I'm using the compaq smart array 5300. Ill give 6.2 a try.

    If i install 6.2 RC1 is it possible to update to stable when it comes out?
    Last edited by Jeremy; 11-22-2006 at 04:57 AM.
    Remote Hands and Your Local Tech for the Los Angeles area.

    (310) 573-8050 - LinkedIn

  9. #9
    Join Date
    Feb 2003
    Location
    Panorama City, CA
    Posts
    2,581

    6.2 blah...

    [bummer]

    So i gave it a go and..
    /dev/da0s1a 507630 37504 429516 8% /

    Start Benchmark Run: Wed Nov 22 04:39:42 PST 2006
    4:39AM up 3 mins, 2 users, load averages: 0.10, 0.10, 0.05

    End Benchmark Run: Wed Nov 22 04:50:10 PST 2006
    4:50AM up 14 mins, 2 users, load averages: 9.65, 4.85, 2.39


    INDEX VALUES
    TEST BASELINE RESULT INDEX

    Dhrystone 2 using register variables 376783.7 9015705.8 239.3
    Double-Precision Whetstone 83.1 1130.5 136.0
    Execl Throughput 188.3 990.5 52.6
    File Copy 1024 bufsize 2000 maxblocks 2672.0 14253.0 53.3
    File Copy 256 bufsize 500 maxblocks 1077.0 9202.0 85.4
    File Read 4096 bufsize 8000 maxblocks 15382.0 158858.0 103.3
    Pipe Throughput 111814.6 1178222.4 105.4
    Pipe-based Context Switching 15448.6 54745.6 35.4
    Process Creation 569.3 2881.9 50.6
    Shell Scripts (8 concurrent) 44.8 279.9 62.5
    System Call Overhead 114433.5 1130366.8 98.8
    =========
    FINAL SCORE 80.7
    6.2-RC1 FreeBSD 6.2-RC1 #0: Thu Nov 16 05:12:08 UTC 2006 root@opus.cse.buffalo.edu:/usr/obj/usr/src/sys/SMP i386

    [/bummer]
    Remote Hands and Your Local Tech for the Los Angeles area.

    (310) 573-8050 - LinkedIn

  10. #10
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    What if you tried disabling SMP? (just for comparison)

  11. #11
    Join Date
    Jun 2003
    Location
    Ukiah, California
    Posts
    394
    Quote Originally Posted by layer0
    What if you tried disabling SMP? (just for comparison)
    I'm still pointing my finger at the disk subsystem.

    Jeremy,

    The link below has some filesystem and kernel tuning tips that may help you out.

    http://groups.google.com/group/maili...fd389f1a99b273
    Jacob Turner -- Ringnebula Systems
    Managed IT solutions for small business

  12. #12
    FreeBSD is known for being significantly slower than it's Linux sisters, however the advantage of a FreeBSD box is that with thousands of visitors to a domain hosted on the BSD box the load will remain low whilst CentOS for example will start crumbling at the knees, freezing up and trying to kernel panic.

    So it's ultimately speed over stability or vice versa.

  13. #13
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    Quote Originally Posted by Jeff - Exceed
    FreeBSD is known for being significantly slower than it's Linux sisters, however the advantage of a FreeBSD box is that with thousands of visitors to a domain hosted on the BSD box the load will remain low whilst CentOS for example will start crumbling at the knees, freezing up and trying to kernel panic.

    So it's ultimately speed over stability or vice versa.
    That is more dependent on the web server used than the OS. I've stress tested CentOs heavily and I don't notice much difference from FreeBSD. But, yes, FreeBSD is more stable (Unix in general is)- but that doesn't mean less load.
    MediaLayer, LLC - www.medialayer.com Learn how we can make your website load faster, translating to better conversion rates for your business!
    The pioneers of optimized web hosting, featuring LiteSpeed Web Server & SSD Storage - Celebrating 10 Years in Business

  14. #14
    Join Date
    Mar 2003
    Location
    California USA
    Posts
    13,681
    Quote Originally Posted by ringnebula
    I'm still pointing my finger at the disk subsystem.

    Jeremy,

    The link below has some filesystem and kernel tuning tips that may help you out.

    http://groups.google.com/group/maili...fd389f1a99b273

    Been there done that on layer0's server did not notice much of an increase.

    Side note, recently had a client that moved from freebsd 6.1 to centos 4 he used the exact same configure lines for apache mysql and php and configurations, i didnt look into it much BUT the loads are cut into half.
    Steven Ciaburri | Industry's Best Server Management - Rack911.com
    Software Auditing - 400+ Vulnerabilities Found - Quote @ https://www.RACK911Labs.com
    Fully Managed Dedicated Servers (Las Vegas, New York City, & Amsterdam) (AS62710)
    FreeBSD & Linux Server Management, Security Auditing, Server Optimization, PCI Compliance

  15. #15
    Join Date
    Mar 2004
    Posts
    1,007
    Has anyone else tried to switch to CentOS or REDHAT and compare them with FreeBSD ?
    Anyone tested it on single core/single cpu system ?

    2x difference in load is a lot.
    Best Regards,
    Namesniper

  16. #16
    Join Date
    Mar 2003
    Location
    California USA
    Posts
    13,681
    Quote Originally Posted by NameSniper
    Has anyone else tried to switch to CentOS or REDHAT and compare them with FreeBSD ?
    Anyone tested it on single core/single cpu system ?

    2x difference in load is a lot.


    Recently we moved an entire cluster to centos (4 webservers, 1 image/memcache server, 1 slave database server, 1 master database server). With the same configuration we haven't had a box lockup. At one point we were using freebsd for all high traffic sites, but recently we have found many of them perform better on centos.
    Steven Ciaburri | Industry's Best Server Management - Rack911.com
    Software Auditing - 400+ Vulnerabilities Found - Quote @ https://www.RACK911Labs.com
    Fully Managed Dedicated Servers (Las Vegas, New York City, & Amsterdam) (AS62710)
    FreeBSD & Linux Server Management, Security Auditing, Server Optimization, PCI Compliance

  17. #17
    Join Date
    Mar 2004
    Posts
    1,007
    Steven can you please keep us posted as you move more servers to centos or other linxu distros ?

    By the way why centos and not other distribution ?
    Best Regards,
    Namesniper

  18. #18
    Join Date
    Mar 2003
    Location
    California USA
    Posts
    13,681
    Quote Originally Posted by NameSniper
    Steven can you please keep us posted as you move more servers to centos or other linxu distros ?

    By the way why centos and not other distribution ?

    Its just the handiest. We don't use any of the prebuilt packages, so its not really a big deal for me.


    I do have a site thats on two servers, one for db and one for web. Its not being moved from freebsd, however when it was on 1 server on centos loads were around 1.80. Its now on two servers and the load on the webserver is always above 2.0. When 6.2 is released I will be upgrading and will let you know what happens.

    Dual core opteron with sata drives btw.
    Steven Ciaburri | Industry's Best Server Management - Rack911.com
    Software Auditing - 400+ Vulnerabilities Found - Quote @ https://www.RACK911Labs.com
    Fully Managed Dedicated Servers (Las Vegas, New York City, & Amsterdam) (AS62710)
    FreeBSD & Linux Server Management, Security Auditing, Server Optimization, PCI Compliance

  19. #19
    Join Date
    Mar 2004
    Posts
    1,007
    I hope that FreeBSD will improve its performance in 6.2 release as beeing 2x times behind CentOS is really not a good idea since perfromance means more hardware and more money for hosting when you can pay less.
    Best Regards,
    Namesniper

  20. #20
    Join Date
    Feb 2003
    Location
    Panorama City, CA
    Posts
    2,581
    INDEX VALUES
    TEST BASELINE RESULT INDEX

    Dhrystone 2 using register variables 376783.7 9012982.0 239.2
    Double-Precision Whetstone 83.1 1132.2 136.2
    Execl Throughput 188.3 997.8 53.0
    File Copy 1024 bufsize 2000 maxblocks 2672.0 25612.0 95.9
    File Copy 256 bufsize 500 maxblocks 1077.0 9434.0 87.6
    File Read 4096 bufsize 8000 maxblocks 15382.0 159547.0 103.7
    Pipe Throughput 111814.6 1154461.4 103.2
    Pipe-based Context Switching 15448.6 80390.6 52.0
    Process Creation 569.3 2865.6 50.3
    Shell Scripts (8 concurrent) 44.8 278.8 62.2
    System Call Overhead 114433.5 1038081.7 90.7
    =========
    FINAL SCORE 87.5
    FreeBSD local.socal.rr.com 6.2-RELEASE FreeBSD 6.2-RELEASE #0: Fri Jan 12 11:05:30 UTC 2007 root@dessler.cse.buffalo.edu:/usr/obj/usr/src/sys/SMP i386



    Same thing when using a single proc. Ill pass on bsd for SMP systems....

    local# dmesg | grep CPU
    CPU: AMD Athlon(tm) MP 2200+ (1792.02-MHz 686-class CPU)
    FreeBSD/SMP: Multiprocessor System Detected: 2 CPUs
    cpu0: <ACPI CPU> on acpi0
    cpu1: <ACPI CPU> on acpi0
    SMP: AP CPU #1 Launched!
    Last edited by Jeremy; 02-05-2007 at 02:06 PM.
    Remote Hands and Your Local Tech for the Los Angeles area.

    (310) 573-8050 - LinkedIn

  21. #21
    Join Date
    Apr 2001
    Location
    Pittsburgh, PA
    Posts
    1,306
    Which benchmark is this? And what are the server specs (dmesg output)?

    Kevin

  22. #22
    Join Date
    Feb 2003
    Location
    Panorama City, CA
    Posts
    2,581
    Hey Kevin,

    In the first page it shows the specs but, dual amd mp 2200+ 2gb raid5.

    dmesg doesn't show anything odd other then a few com port errors

    but running centos or fedora she scores in the 120's and have no performance issues.

    you might know more about freebsd then i do, i can give you a login ( this box is at my house before production use

    As for the benchmark its one in the Dedicated forum VPS vs Dedicated servers, i used it just to see if i was going nuts when i would compile (ports or from source) or untar files everything was taking ages to complete.
    Remote Hands and Your Local Tech for the Los Angeles area.

    (310) 573-8050 - LinkedIn

  23. #23
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    Quote Originally Posted by sigma
    Which benchmark is this? And what are the server specs (dmesg output)?

    Kevin
    I know you have a lot of BSD-based servers, is there any chance that you can use this benchmark on one of your own?

  24. #24
    What benchmark is it?

  25. #25
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    wget/fetch http://members.dslextreme.com/users/...1.0-wht.tar.gz

    gunzip -dvc unixbench-4.1.0-wht.tar.gz | tar xvf -

    cd unixbench-4.1.0-wht

    make

    ./Run

Page 1 of 2 12 LastLast

Posting Permissions

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