Litespeed Vs. Apache web server - Real world page loading speed test results

Discussion in 'Computer Science | IT | Networking' started by Kaustubh Katdare, Oct 2, 2013.

  1. Kaustubh Katdare

    Administrator

    Engineering Discipline:
    Electrical
    So a few days ago, we asked all our members whether the site was loading slower for all of you and almost everyone said yes. Upon inspection, I discovered that the pages were being served by Apache web server and not Litespeed - which we had configured. It so happened that a license issue of Litespeed caused the webserver go down and automatically switch to Apache. As a result, the pages were loading slower and the ajax requests (for example @user tagging) were being served with an annoying delay. Though we were unaware of the problem, we didn't have graphs or stats to discover what's the overall impact.

    Thankfully, Google Webmaster Tools (GWT) came to our rescue, exactly showing how Google viewed CrazyEngineers. Check the graph which I just obtained from GWT.

    crazyengineers.

    The hills you see in the graph were created when the site was running on Apache Web Server. After correcting the license issue, the page loading times were back to normal. We've all the reasons to believe that average page loading speeds of about ~200 - 250 milliseconds is extremely fast for a site like ours (lot of content, users and images).

    LiteSpeed is a drop-alternative to Apache web server, which apparently is the most popular web server. Note that we're running a commercial version of the server as against the free software - Apache. I've heard that webmasters have noticed similar improvements with NGIX; but I've no tests or access to test machines to check it all by myself.

    If anyone of you has had a chance to play with different web servers, please let me know your observations. I'm curious to know what exactly Litespeed does differently that it gains that much of an edge over apache?
     
    • Like Like x 1
    #1
  2. avii

    avii Ace

    Engineering Discipline:
    Computer Science
    #2
  3. lsmichael

    lsmichael Newbie

    Engineering Discipline:
    Computer Science
    Howdy,

    LiteSpeed Technologies here. There are a number of differences between LiteSpeed Web Server (LSWS) and Apache. The largest difference is LSWS's event-driven architecture, which means that it handles multiple connections much better, using less CPU and RAM, and pulls out static content much faster. A page like this, using Xenforo, which is written in PHP, will also see some performance benefits from our use of LSAPI.

    Hope that helps clear things up. Glad LSWS is helping you so much.

    Cheers,

    Michael
     
    • Like Like x 2
    #3
  4. Kaustubh Katdare

    Administrator

    Engineering Discipline:
    Electrical
    @lsmichael - It's a pleasure to have you with us on CE. I can confirm that Litespeed handles load much better than Apache; we've had record number of simultaneous users online a few days ago and LSWS didn't acknowledge that there was load. Apache would have gone down for similar loads; as it did in the past.

    Would love to know what exactly does LSAPI offer which result into faster PHP performance?
     
    #4
  5. Anand Tamariya

    Engineering Discipline:
    IT
    I think the performance of LSAPI is good only for PHP because it focuses on that. Apache on the other hand is a generic webserver and works for variety of use-cases given one understands how to tune it properly.
     
    #5
  6. lsmichael

    lsmichael Newbie

    Engineering Discipline:
    Computer Science
    Howdy again.

    LSAPI provides better-than-Apache performance for PHP, Ruby, and now Python: http://www.litespeedtech.com/products/litespeed-sapi/overview

    We do have the longest history working with PHP and PHP LSAPI has the most features of all the LSAPI modules.

    It is true that in some (many?) cases Apache optimization can achieve results similar to LSWS (especially if you're willing to buy hardware to make up for Apache's CPU and memory usage issues). But I feel like many people, especially experienced sysadmins, like to say what you've said — that you can tune Apache to do what LSWS does. That statement, I feel, ignores the fact that tuning Apache is not a simple task. It may be possible, but it is time-consuming, even for experienced sysadmins, and expensive if you have to pay someone else to do it.

    Most of the difference is actually just code optimization. In different environments (like shared hosting) the extra features, like suEXEC daemon mode which then allows for effective opcode caching, can also make a big difference.
     
    • Like Like x 1
    #6
: PHP

Share This Page