Results 1 to 9 of 9
  1. #1

    Help on server performance

    We currently run two servers (UNIX Intel P4 2.53 GHz) with a load balance addition. Whilst we’ve been told this is more than suitable for our needs, we are finding this is not the case. We use the servers for simple swf file downloads and nothing else, a typical file will vary in size from 150k to 750k. Speed of initial file download start is business critical. We tend to have high volume of file call ups in peaks of traffic.

    When we are running at 10Mb/s across the two servers everything is fine – a file will start to download in 1 second on broadband and 5 seconds on dial-up. But when we have a heavy day this will peak at 60-80 Mb/s across the two servers and file call up slows down (5 & 15 seconds respectfully) and we get interrupted file downloads. We are told the servers could cope with 120 Mb/s combined, but at half this we find the servers are struggling. Have we got the right set up? Do we need more powerful servers? Do we need more (loathed to do this as our business could require endless servers as it grows)? Are we right to use dedicated servers – should we be on a distributed server network?

  2. #2
    The most important is the tuning of the TCP stack and the webserver if you are serving static content. If you are running a connection tracking firewall, make sure that the maximum number of connections is set to a very high number.

    If you can provide more info, ie. OS, webserver, etc, I can provide you with more help if needed.
    ::. www.diginode.net : Dedicated Servers : Virtual Machine Servers .::
    ::. Industry-Leading Remote Server Management .::
    ::. Automatic OS Re-image : Instant Server Reboot : Remote Serial Console .::
    ::. Over 20 OSes to choose from : Install a new OS every day .::

  3. #3
    We serve video as swf files into online ads. So volumes tend to peak at certain times of the day. When one of our ads is served on page it initiates a call to the servers to start the video swf download - speed is therefore important. I'm not a server expert (by a long way) OS is BSD?

  4. #4
    definitely should switch from apache to kernel-mode webserver if you have not done so yet, you'll see like 5-10x performance boost.

  5. #5
    flashtalking,

    your bottleneck is inefficient pagecache and apache, the latter being the bigger problem. you need to:

    * ensure that your most commonly accessed dataset is in pagecache at all times (if you are using freebsd, you are going to have issues because of no dcache)
    * use a single-threaded non-blocking i/o (async i/o even better) httpd. you are not serving dynamic content (ie php), so ditch the apache with its horrible bloat.
    * use linux 2.6.4 and up. benchmarks show that 2.6 is the only kernel (comparison between 2.4, 2.6 fbsd and nbsd) that maintains O(1) complexity when scaling to a large number of http requests.
    * take a long hard look at your load balancer, it may have issues keeping up. if you use a hardware load balancer acting as a proxy, i would suggest switching to weighted round robin dns, since you do not need to maintain session affinity and adding a proxying device is tacking latency onto each transaction.

    we've built a lot of similar clusters, so you are welcome to get in touch with me off the boards if you want advice that is better geared towards your particular situation.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  6. #6
    Thanks Paul
    It's also been suggested our connection may be too slow but we use 100Mb/s NIC cards on each server and each port, so should I discount this?
    Flashtalking (aka Paul as well ;-))

  7. #7
    Originally posted by rusko

    * use linux 2.6.4 and up. benchmarks show that 2.6 is the only kernel (comparison between 2.4, 2.6 fbsd and nbsd) that maintains O(1) complexity when scaling to a large number of http requests.
    You have any link of this benchmark?

  8. #8
    Originally posted by flashtalking
    Thanks Paul
    It's also been suggested our connection may be too slow but we use 100Mb/s NIC cards on each server and each port, so should I discount this?
    Flashtalking (aka Paul as well ;-))
    paul,

    generally, when you get to about 50 megs on a faste nic you want to upgrade to gige. with that said, this is not the problem in your setup imo, unless you are seeing packet loss at the same time you are seeing the slowdown.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  9. #9
    Originally posted by icehosting
    You have any link of this benchmark?
    i thought i did, but i do not. google for it, it was also on /. i think.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

Posting Permissions

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