Page 3 of 7 FirstFirst 123456 ... LastLast
Results 81 to 120 of 267
  1. #81
    Join Date
    May 2006
    Location
    Iowa
    Posts
    2,612
    I wonder if you will see hosts in the future saying they have varnish like the hosts we have now that advertise saying they have lightspeed.
      0 Not allowed!

  2. #82
    Join Date
    Feb 2008
    Location
    Houston, Texas, USA
    Posts
    2,955
    Quote Originally Posted by Dan_EZPZ View Post
    One client found it wouldn't even work with their domain, cpanel******.com was just being thrown to a cPanel login page and not showing their site. I'd imagine its the same if your domain happens to begin with whm.
    I can confirm this bug was fixed in 1.4.5rc.

    The bug is that URL opt out regex "cpanel." needed an escape of the dot like this: "cpanel\.". Otherwise it matches both cpanelmydomain.tld and cpanel.mydomain.tld. cpanel.mydomain.tld must be routed to cPanel whereas cpanelmydomain.tld to a specific Apache vhost.

    If you want this fixed for a pre-1.4.5rc release, simply open /etc/varnish/cpanel.url.vcl and escape the dot like this:

    from
    if (req.http.host ~ "^cpanel.")
    to
    if (req.http.host ~ "^cpanel\.")

    Restart Varnish and you're good.

    Regards
    Last edited by UNIXy; 04-05-2011 at 02:05 PM.
    UNIXy - Fully Managed Servers and Clusters - Established in 2006
    [ cPanel Varnish Nginx Plugin ] - Enhance LiteSpeed and Apache Performance
    www.unixy.net - Los Angeles | Houston | Atlanta | Rotterdam
    Love to help pro bono (time permitting). joe > unixy.net
      0 Not allowed!

  3. #83
    Join Date
    May 2010
    Posts
    613
    i have to say i tried litespeed and for me the ram usage was great but the processor usage wasn't much greater than apache. however we went back to apache (we where using apache and nginx) and varnish (i had tried varnish but became concerned about how it used ram as a cache, incase it used all the ram and borught down the whole server. i did google to find out how it worked but couldn't find any meaningfull information)

    however we have it running on our server and it's amazing it uses a 1G chunk of ram for caching and it writes out old cache data to the hard drive (so i'm told) but even if it just drops old data that's cool.

    performance is great and processor usage is low although ram usage is higher than both litespeed and nginx, but that's how varnish caches.

    joe
    www.JGwebhosting.co.uk - CPanel Control panel, domain registration, Reseller packages And 24X7 Technical support! - Now Taking in Hosting Refugee's

    www.JGpcrepair.co.uk - Computer repair, New Systems, Software, Hardware
      0 Not allowed!

  4. #84
    Join Date
    Feb 2008
    Location
    Houston, Texas, USA
    Posts
    2,955
    I just wanted to add that almost all reported bugs are fixed within the hour. I can't locate a bug report from you on this on this. So don't hesitate to report any issue you're facing. So far with 1.4.5rc there has been one feature request and no bug report.

    Best
    UNIXy - Fully Managed Servers and Clusters - Established in 2006
    [ cPanel Varnish Nginx Plugin ] - Enhance LiteSpeed and Apache Performance
    www.unixy.net - Los Angeles | Houston | Atlanta | Rotterdam
    Love to help pro bono (time permitting). joe > unixy.net
      0 Not allowed!

  5. #85
    The biggest concern with using varnish or any reverse proxy in front of Apache for shared hosting server is accuracy of bandwidth usages. Bandwidth usage of all requests that are served from Varnish cache without hitting the backend web server is not counted by cPanel bandwidth usage statistics. If your hosting plan has a bandwidth limit, you may lose money if the bandwidth usage is over the limit but cPanel show the bandwidth usage is still under.
    LiteSpeed always log all requests in to access and bandwidth log files no matter cache is in use or not.

    As to performance, for serving a page out of cache, LiteSpeed Cache is about 10-50% faster than varnish, we have done benchmark comparisons on this.

    When compare performance of a cached page to a LiteSpeed dynamically generated page, the cached page will always won, no matter it is varnish, squid or Apache mod_cache.

    So, the overall performance improvement is mainly decided by cache hit rate, higher hit rate will result in better observed performance.

    Traditionally, a cache proxy only cache pages that does not change from visitor to visitor, we could call it "public cache". Not all pages are "public cache" friendly, especially for dynamic pages.

    Varnish VCL is flexible, it can be configured to cache a response for each individual visitor, and we could call it "private cache". That’s the cache policy used by UNIXy plug-in. The advantage of "private cache" is to be able cache a page that has response header "Cache-Control: private, no-cache, no-store" for a short period of time without causing much trouble. One visitor will not be accidentally served with a page cached for another user. The disadvantage of "private cache" is that it will reduce the cache hit rate and increases cache storage usage if a page actually is "public cache" friendly.

    Currently, LiteSpeed cache is mainly used for "public cache" friendly pages, we have published rewrite rules for enabling cache for a few most popular web applications, like vBulletin, WordPress, Joomla, Mediawiki, drupal, etc.

    For "private cache", we plan to add that after our 4.1 release this week. The implementation is super easy, just add client IP and cookie as part of hash key for store and lookup cached page, so it should be available in a week or two. Then user can switch between "public cache" and "private cache" policy to provide maximum performance.

    Cache is a great way to improve performance, however, it has its own shortcomings. The biggest one is a cached page becomes stale if it has been updated before it expires. You may need to take extra time to answer questions like why my page or image does not update? :-)

    Talking about "flexibilities", even though, Varnish VCL is powerful and flexible, it is stored in a central place, end user could not change cache policy at all once set by administrator, and it is too complicated for end users even average web master to handle. However, with LiteSpeed cache, end user can change and fine tune cache policy with familiar rewrite rules through .htaccess, to use "public cache", "private cache" or to turn off cache completely at URL level.

    Just want to give everyone a head up about caching from technical point of view, if anything wrong, please correct me. :-)
    LiteSpeed Web Acceleration Platform by http://www.litespeedtech.com
    Apache drop-in replacement. Triple server capacity with 10X performance increases.
    Ultimate web serving platform for WordPress, Magento and other web applications.
    Turbo charging all WordPress sites hosted on your server with a single click!
      0 Not allowed!

  6. #86
    Join Date
    Feb 2008
    Location
    Houston, Texas, USA
    Posts
    2,955
    Hi Mistwang,

    I'd like to comment on this since you mentioned UNIXY's plugin. Keep in mind that I've already commented on all of these concerns (read: lack of understanding of Varnish) in the main thread and showed that 90% of these "concerns" are due to your misunderstanding and inexperience with Varnish and its tools (varnishncsa, varnishlog, varnishstat, varnishtop, etc). Please read up on those.

    Quote Originally Posted by mistwang View Post

    As to performance, for serving a page out of cache, LiteSpeed Cache is about 10-50% faster than varnish, we have done benchmark comparisons on this.

    When compare performance of a cached page to a LiteSpeed dynamically generated page, the cached page will always won, no matter it is varnish, squid or Apache mod_cache.

    So, the overall performance improvement is mainly decided by cache hit rate, higher hit rate will result in better observed performance.

    Traditionally, a cache proxy only cache pages that does not change from visitor to visitor, we could call it "public cache". Not all pages are "public cache" friendly, especially for dynamic pages.

    Varnish VCL is flexible, it can be configured to cache a response for each individual visitor, and we could call it "private cache". That’s the cache policy used by UNIXy plug-in. The advantage of "private cache" is to be able cache a page that has response header "Cache-Control: private, no-cache, no-store" for a short period of time without causing much trouble. One visitor will not be accidentally served with a page cached for another user.
    I'm trying to understand this. Are you explaining why Varnish is faster than Litespeed?

    Quote Originally Posted by mistwang View Post
    The disadvantage of "private cache" is that it will reduce the cache hit rate and increases cache storage usage if a page actually is "public cache" friendly.
    Private cache is an advantage not a disadvantage. That's the improvement many are reporting.

    Quote Originally Posted by mistwang View Post
    For "private cache", we plan to add that after our 4.1 release this week. The implementation is super easy, just add client IP and cookie as part of hash key for store and lookup cached page, so it should be available in a week or two. Then user can switch between "public cache" and "private cache" policy to provide maximum performance.
    I thought you just said it's a disadvantage. Why are you implementing it in Litespeed then? Anyway, I'm glad to see you're implementing this approach to caching in the next release to improve performance. That's what our plugin has been doing since day one.

    Quote Originally Posted by mistwang View Post
    Cache is a great way to improve performance, however, it has its own shortcomings. The biggest one is a cached page becomes stale if it has been updated before it expires. You may need to take extra time to answer questions like why my page or image does not update? :-)
    There's a work around with Varnish. You could exclude a URL, a vhost, or just add "dev\." to the URL exclusion list for people that need to do web development work and not need performance while in that stage.

    Quote Originally Posted by mistwang View Post
    Talking about "flexibilities", even though, Varnish VCL is powerful and flexible, it is stored in a central place, end user could not change cache policy at all once set by administrator, and it is too complicated for end users even average web master to handle. However, with LiteSpeed cache, end user can change and fine tune cache policy with familiar rewrite rules through .htaccess, to use "public cache", "private cache" or to turn off cache completely at URL level.
    You're just not familiar with the VCL enough. All the above is possible. Yes, you can have a user change TTL on a per-sub/domain level. It doesn't even require a Varnish restart.

    Quote Originally Posted by mistwang View Post
    Just want to give everyone a head up about caching from technical point of view, if anything wrong, please correct me. :-)
    I'm just disappointed you're not informing yourself about Varnish considering it has much more advanced technology than Litespeed. In terms of performance, you could've just summed up your comment here by saying: Litespeed will soon have what Varnish has had for many years now.

    Regards
    UNIXy - Fully Managed Servers and Clusters - Established in 2006
    [ cPanel Varnish Nginx Plugin ] - Enhance LiteSpeed and Apache Performance
    www.unixy.net - Los Angeles | Houston | Atlanta | Rotterdam
    Love to help pro bono (time permitting). joe > unixy.net
      0 Not allowed!

  7. #87
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,574
    Unixy,

    I think the main point here is that LiteSpeed is a web server product first, not a caching product. So Varnish vs LiteSpeed is somewhat of an apples to oranges comparison in itself. It seems LiteSpeed will be implementing caching features to compete with some of the advantages that Varnish brings, which is great. But the comparison of a dynamically generated page (from any web server) to a cached page seems unfair to begin with. Yes I understand that you're just comparing the end result of using one solution vs another, but the correct comparison here is really just LiteSpeed Cache vs Varnish, not LiteSpeed Web Server (or any other web server) vs Varnish.
      0 Not allowed!

  8. #88
    Join Date
    Jul 2005
    Location
    In the Internets
    Posts
    3,611
    Varnish is decent, it'll be good for people who can't afford Litespeed but want better performance of Apache.
      0 Not allowed!

  9. #89
    Join Date
    Nov 2002
    Location
    United Arab Emirates
    Posts
    467
    If you guys check the Title of the thread. I named it " Litespeed vs Apache + Varnish", meaning comparison between Litespeed and (Apache + Varnish). Litespeed is a webserver and so is Apache.

    And here is my initial question in my First Post :
    Which server would give the fatest loading times for a typical CMS.
    A server running on Litespeed or Apache optimized by Varnish ?
    I am asking specifically from an End User perspective, what the end user needs is the optimal speed along with readability.

    Now what happened was that this thread got hijacked by the hosting companies trying to prove Varnish is better than Litespeed, why was Apache out of the picture ?


    I really dont care for the bickering of the hosting companies among themselves, this thread was for me, from an end user perspective.

    I will choose the route which delivers the best possible speed + reliability.


    Stop the bickering
      0 Not allowed!

  10. #90
    Join Date
    Feb 2008
    Location
    Houston, Texas, USA
    Posts
    2,955
    Quote Originally Posted by layer0 View Post
    Unixy,

    I think the main point here is that LiteSpeed is a web server product first, not a caching product. So Varnish vs LiteSpeed is somewhat of an apples to oranges comparison in itself. It seems LiteSpeed will be implementing caching features to compete with some of the advantages that Varnish brings, which is great. But the comparison of a dynamically generated page (from any web server) to a cached page seems unfair to begin with. Yes I understand that you're just comparing the end result of using one solution vs another, but the correct comparison here is really just LiteSpeed Cache vs Varnish, not LiteSpeed Web Server (or any other web server) vs Varnish.
    I agree with this.

    Regards
    UNIXy - Fully Managed Servers and Clusters - Established in 2006
    [ cPanel Varnish Nginx Plugin ] - Enhance LiteSpeed and Apache Performance
    www.unixy.net - Los Angeles | Houston | Atlanta | Rotterdam
    Love to help pro bono (time permitting). joe > unixy.net
      0 Not allowed!

  11. #91
    Joe/UNIXy,

    I am not going to continue your meaning less arguments. My time can be better spent than doing that. :-)

    But I have to correct your error from technical point of view.
    Private cache is an advantage not a disadvantage.
    It depends, as I said, if a page is "public cache" friendly, "private cache" may reduces cache hit rate, but it better than no cache.
    If a page is not "public cache" friendly, only "private cache" can be used, it is an advantage, better than no cache.
    The best cache practice is to apply "public cache" for friendly pages, use "private cache" for non-friendly pages if stale cache page is not a big deal.
    LiteSpeed Web Acceleration Platform by http://www.litespeedtech.com
    Apache drop-in replacement. Triple server capacity with 10X performance increases.
    Ultimate web serving platform for WordPress, Magento and other web applications.
    Turbo charging all WordPress sites hosted on your server with a single click!
      0 Not allowed!

  12. #92
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
    Quote Originally Posted by mistwang View Post
    The biggest concern with using varnish or any reverse proxy in front of Apache for shared hosting server is accuracy of bandwidth usages. Bandwidth usage of all requests that are served from Varnish cache without hitting the backend web server is not counted by cPanel bandwidth usage statistics. If your hosting plan has a bandwidth limit, you may lose money if the bandwidth usage is over the limit but cPanel show the bandwidth usage is still under.
    LiteSpeed always log all requests in to access and bandwidth log files no matter cache is in use or not.
    This is 100% incorrect. We run all our servers and IPs monitored by Cacti and I can verify that cPanel with Varnish installed reports the correct bandwidth usage. Please stop spreading FUD when you do not have data to backup these claims.
      0 Not allowed!

  13. #93
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,574
    Quote Originally Posted by brentpresley View Post
    This is 100% incorrect. We run all our servers and IPs monitored by Cacti and I can verify that cPanel with Varnish installed reports the correct bandwidth usage. Please stop spreading FUD when you do not have data to backup these claims.
    With all due respect, I don't know if that's FUD. His explanation is quite logical, unless UNIXy has specifically implemented something that can count the bandwidth usage properly. Otherwise, it would course be inaccurate, to some extent (depends on cache rate).
      0 Not allowed!

  14. #94
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
      0 Not allowed!

  15. #95
    Quote Originally Posted by Farrukh View Post
    I really dont care for the bickering of the hosting companies among themselves, this thread was for me, from an end user perspective.

    I will choose the route which delivers the best possible speed + reliability.
    OK, I will try. But there is no definitive answer either, many other factors are involved if you compare two companies, like hardware, accounts hosted on one server, level of support,etc.

    If only difference is the software used, you can think in this way.
    Let me start with something that most likely everyone agrees,

    LiteSpeed (no cache) > Plain Apache

    LiteSpeed (no cache) + Varnish > Plain Apache + Varnish

    What about LiteSpeed + internal Cache comparing to Apache + Varnish?
    that's where the debates starts. You probably do not want to hear more about it.

    Let's follow the logic above, assuming LiteSpeed cache itself is as fast as Varnish (our benchmarks show LiteSpeed cache is actually faster, believe it or not), if LiteSpeed and Varnish uses the same cache policy and have the same cache hit rates, the conclusion is

    LiteSpeed (with cache) = LiteSpeed (no cache) + Varnish > Plain Apache + Varnish

    I think someone may start to argue about my assumption anyway.

    What about Litespeed (no cache) vs Apache + Varnish?

    It is likely that Apache + Varnish is faster for PHP pages, for serving static files, LiteSpeed is still faster.

    Hosting companies using LiteSpeed are likely to embrace public cache + private cache, once we make private cache feature available in a few weeks.

    Even without "private cache" today, "public cache" policy could be applied towards pages from popular CMS, but that need manual configurations.

    I hope it helps.
    LiteSpeed Web Acceleration Platform by http://www.litespeedtech.com
    Apache drop-in replacement. Triple server capacity with 10X performance increases.
    Ultimate web serving platform for WordPress, Magento and other web applications.
    Turbo charging all WordPress sites hosted on your server with a single click!
      0 Not allowed!

  16. #96
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
    From our experience nothing, including LiteSpeed Cache 2-CPU version, is as fast as the following:

    Apache + mod_ruid2 + eAccelerator + Varnish.

    With that setup, you have GREAT user-instance security (mod_ruid2), GREAT PHP generation (mod_ruid2), GREAT PHP acceleration for generated PHP compiled files (eAccelerator), and GREAT static+dynamic user and public memory caching.

    Toss in a GOOD firewall and router-level security, and there is not a lot else you can ask for.
      0 Not allowed!

  17. #97
    Quote Originally Posted by brentpresley View Post
    Might "make sense", but doesn't match what we see.
    It is easy to verify, just create a test account. put a static file there.
    use "ab" to hammer that file, check access log and bandwidth log for that domain under /usr/local/apache/domlogs/, it is likely only one or a few log entries in there, far less than the number of requests in the "ab" run. Number of log entries is depends on the cache expire time and how long the ab test last.
    LiteSpeed Web Acceleration Platform by http://www.litespeedtech.com
    Apache drop-in replacement. Triple server capacity with 10X performance increases.
    Ultimate web serving platform for WordPress, Magento and other web applications.
    Turbo charging all WordPress sites hosted on your server with a single click!
      0 Not allowed!

  18. #98
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
    Sorry dude, already been down this route. And confirmed it on the router level.

    Since running Varnish cache for a month on one of our shared public servers, the traffic reported by the monitoring server (Cacti), the router (Cisco), and the cPanel application are all the same.


    But if you have data to prove otherwise, we would all appreciate that you post it.
      0 Not allowed!

  19. #99
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,574
    Quote Originally Posted by brentpresley View Post
    Sorry dude, already been down this route. And confirmed it on the router level.

    Since running Varnish cache for a month on one of our shared public servers, the traffic reported by the monitoring server (Cacti), the router (Cisco), and the cPanel application are all the same.


    But if you have data to prove otherwise, we would all appreciate that you post it.
    Without even doing any testing, what you're saying logically makes no sense, UNLESS UNIXy has developed their own bandwidth recording system for Varnish that integrates with cPanel.

    You need to relax a bit Brent - mistwang is raising some valid points over the discussion here. I'm glad you found a setup that you think delivers the best performance, we got that, so would appreciate if you gave it a rest.
      0 Not allowed!

  20. #100
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
    Alright, why don't we put this to a test, and end the dispute once and for all.

    I will donate a VDS with up to 4 cores and 8GB of RAM for testing of LiteSpeed vs. UNIXy Varnish.

    Prior to the test, I need someone to select a grouping of websites to be loaded onto the server as a "representative sampling" of current-day dynamic and static web content. A fair mix, so to speak, of what users actually USE in 2011.

    I will give each party 1 week to install, configure, and optimize their setup prior to testing.

    For testing purposes, AppLogic provides a built-in load generator that can max out just about any server I have EVER seen, and can be customized to use as many URLs as we want. It gives statistics back on the response rate from the server.

    Any other agreed-upon benchmark utilities can also be run.



    Does this sound fair to everyone?
      0 Not allowed!

  21. #101
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,574
    Brent,

    I think mistwang's post above is already really clear. LiteSpeed is refining their cache to allow for what's referred to as a "public cache" which will make their caching system as good or better than what's offered with Varnish through customized VCLs.

    It's great that you want to do this testing but I think the answer here is already pretty clear. You need to compare the right things, that's all. Comparing any web server without a cache to a web server with a cache is obviously going to be skewed.
      0 Not allowed!

  22. #102
    Join Date
    Feb 2008
    Location
    Houston, Texas, USA
    Posts
    2,955
    Do you not know about varnishncsa? These logs are produced by Varnish. These are cache hits and are served directly from the cache:

    # /usr/local/varnish/bin/varnishncsa
    x.x.x.x - - [06/Apr/2011:04:04:28 +0400] "GET http://www.fastlayer.net/ HTTP/1.0" 200 9 "-" "Lynx/2.8.5rel.1 libwww-FM/2.14 SSL-MM/1.4.1 OpenSSL/0.9.8e-fips-rhel5"
    x.x.x.x - - [06/Apr/2011:04:04:35 +0400] "GET http://www.fastlayer.net/ HTTP/1.0" 200 9 "http://www.fastlayer.net/" "Lynx/2.8.5rel.1 libwww-FM/2.14 SSL-MM/1.4.1 OpenSSL/0.9.8e-fips-rhel5"
    x.x.x.x - - [06/Apr/2011:04:04:37 +0400] "GET http://www.fastlayer.net/ HTTP/1.0" 200 9 "http://www.fastlayer.net/" "Lynx/2.8.5rel.1 libwww-FM/2.14 SSL-MM/1.4.1 OpenSSL/0.9.8e-fips-rhel5"

    I thought this thread was about performance (lsws vs. varnish).

    Regards
    UNIXy - Fully Managed Servers and Clusters - Established in 2006
    [ cPanel Varnish Nginx Plugin ] - Enhance LiteSpeed and Apache Performance
    www.unixy.net - Los Angeles | Houston | Atlanta | Rotterdam
    Love to help pro bono (time permitting). joe > unixy.net
      0 Not allowed!

  23. #103
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
    Quote Originally Posted by layer0 View Post
    Brent,

    I think mistwang's post above is already really clear. LiteSpeed is refining their cache to allow for what's referred to as a "public cache" which will make their caching system as good or better than what's offered with Varnish through customized VCLs.

    It's great that you want to do this testing but I think the answer here is already pretty clear. You need to compare the right things, that's all. Comparing any web server without a cache to a web server with a cache is obviously going to be skewed.
    So I am offering 3rd party resources, free, for an apples to apples comparison and you are saying this is not worthwhile?


    Remember, I did say apples to apples, as close as we can get it. Caching vs. caching. Tweaked server vs. tweaked server.
      0 Not allowed!

  24. #104
    Join Date
    Nov 2002
    Location
    United Arab Emirates
    Posts
    467
    Quote Originally Posted by mistwang View Post
    OK, I will try. But there is no definitive answer either, many other factors are involved if you compare two companies, like hardware, accounts hosted on one server, level of support,etc.

    If only difference is the software used, you can think in this way.
    Let me start with something that most likely everyone agrees,

    LiteSpeed (no cache) > Plain Apache

    LiteSpeed (no cache) + Varnish > Plain Apache + Varnish

    What about LiteSpeed + internal Cache comparing to Apache + Varnish?
    that's where the debates starts. You probably do not want to hear more about it.

    Let's follow the logic above, assuming LiteSpeed cache itself is as fast as Varnish (our benchmarks show LiteSpeed cache is actually faster, believe it or not), if LiteSpeed and Varnish uses the same cache policy and have the same cache hit rates, the conclusion is

    LiteSpeed (with cache) = LiteSpeed (no cache) + Varnish > Plain Apache + Varnish

    I think someone may start to argue about my assumption anyway.

    What about Litespeed (no cache) vs Apache + Varnish?

    It is likely that Apache + Varnish is faster for PHP pages, for serving static files, LiteSpeed is still faster.

    Hosting companies using LiteSpeed are likely to embrace public cache + private cache, once we make private cache feature available in a few weeks.

    Even without "private cache" today, "public cache" policy could be applied towards pages from popular CMS, but that need manual configurations.

    I hope it helps.

    @mistwang, Thanks this helps a lot.

    One more question, my initial inquiry was about running a CMS, a PHP CMS which would serve dynamic pages. Which setup would be the best ?
      0 Not allowed!

  25. #105
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
    Quote Originally Posted by Farrukh View Post
    @mistwang, Thanks this helps a lot.

    One more question, my initial inquiry was about running a CMS, a PHP CMS which would serve dynamic pages. Which setup would be the best ?
    Farrukh, these are all good questions, just get several opinions. Obviously UNIXy is the author of his Varnish plugin. And mistwang is one of the developers (IIRC) for LiteSpeed. Just so all that is in the open and you know.
      0 Not allowed!

  26. #106
    Join Date
    Jul 2005
    Location
    In the Internets
    Posts
    3,611
    Quote Originally Posted by brentpresley View Post
    Alright, why don't we put this to a test, and end the dispute once and for all.

    I will donate a VDS with up to 4 cores and 8GB of RAM for testing of LiteSpeed vs. UNIXy Varnish.

    Prior to the test, I need someone to select a grouping of websites to be loaded onto the server as a "representative sampling" of current-day dynamic and static web content. A fair mix, so to speak, of what users actually USE in 2011.

    I will give each party 1 week to install, configure, and optimize their setup prior to testing.

    For testing purposes, AppLogic provides a built-in load generator that can max out just about any server I have EVER seen, and can be customized to use as many URLs as we want. It gives statistics back on the response rate from the server.

    Any other agreed-upon benchmark utilities can also be run.



    Does this sound fair to everyone?
    It was already done:

    http://www.litespeedtech.com/support...hlight=varnish
      0 Not allowed!

  27. #107
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
    Quote Originally Posted by nerdie View Post
    So did you actually read the test page?

    The "test" page was a static PHP page (Document Path: /phpinfo.php
    Document Length: 39265 bytes).

    Completely static page that is not exactly representative of dynamic content.


    Furthermore, the testing was ONLY done by LiteSpeed staff. Not implying anything, but if you want the results to be viewed as fair and unbiased, they cannot be the ones doing the testing.


    Thirdly, they did not tell us the details of their server setup. And better yet, give us the Varnish VCL file to examine for ourselves. The configuration of the server tested is pretty much left for guessing, both for the LiteSpeed and the Varnish runs (how many cores, how much RAM allocated to cache, etc. etc.). They state that the dedicated server listed in a previous post could not be obtained, so they used a VDS, but aside from the GB of RAM (6), they did not give us other details like core count. And which version of LiteSpeed (how many cores?) they used.



    So after all that, I would say my offer for some REAL testing is still valid and on the table. As a static PHP page only represents a fraction of the content out there today.
      0 Not allowed!

  28. #108
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,574
    Quote Originally Posted by brentpresley View Post
    Furthermore, the testing was ONLY done by LiteSpeed staff. Not implying anything, but if you want the results to be viewed as fair and unbiased, they cannot be the ones doing the testing.
    The user who posted in that thread doesn't appear to work for LiteSpeed.
      0 Not allowed!

  29. #109
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
    Quote Originally Posted by layer0 View Post
    The user who posted in that thread doesn't appear to work for LiteSpeed.
    Oh really?

    Since no 3rd party's have stepped up to the mark of doing this testing someone from Litespeed will kindly kick this public show off, Once the testing is underway am sure people will step up, Remember only trusted people will be allowed to undertake the testing.
      0 Not allowed!

  30. #110
    Quote Originally Posted by Farrukh View Post
    @mistwang, Thanks this helps a lot.

    One more question, my initial inquiry was about running a CMS, a PHP CMS which would serve dynamic pages. Which setup would be the best ?
    When access one URL, if all visitors get exactly the same content, using "public cache" policy is the best choice. if it shows slightly different content for different visitor, "private cache" policy can be used.
    Varnish use "private cache" only for dynamic page ( I maybe wrong on this though), LiteSpeed will be able to use either "private" or "public" cache, your choice.

    Usually, from end user point of view, visitors may not feel much difference when pages are served from cache, no matter it is varnish or litespeed, as long as it is cached.

    Again, you need consider other factors, not just software used, it does not necessarily guarantee you good service.
    LiteSpeed Web Acceleration Platform by http://www.litespeedtech.com
    Apache drop-in replacement. Triple server capacity with 10X performance increases.
    Ultimate web serving platform for WordPress, Magento and other web applications.
    Turbo charging all WordPress sites hosted on your server with a single click!
      0 Not allowed!

  31. #111
    Join Date
    Feb 2008
    Location
    Houston, Texas, USA
    Posts
    2,955
    Quote Originally Posted by nerdie View Post
    The title of the thread: LiteSpeed VS Unixy Varnish - The Truth!

    That's a lie!

    Whoever is running that show did not benchmark our plugin.

    They downloaded the default VCL from the Varnish repository, which everyone knows is suboptimal at best, and slapped the "Unixy" tag on it. Run the benchmark against the UNIXy plugin and get me involved. See I'm personally not tied to any one Web server or technology. Our clients run just all sorts of configurations from one VPS to clusters. Yes, some even run Litespeed!

    I don't mind accepting the fact that Litespeed could possibly run faster (if it were to be proven). However, Litespeed Tech will never accept the fact that Varnish has performed faster. Someone in this forum ran benchmarks (a Litespeed Tech client mind you) and found Varnish to be 500 req/s faster than Litespeed Cache!

    You'd think Litespeed Tech would stop misleading people with these lies. You know it's easy to tell from the benchmark output that this is not UNIXy's plugin, right? But you have the SEO upper arm right now so sadly folks will keep being misled.

    Regards
    UNIXy - Fully Managed Servers and Clusters - Established in 2006
    [ cPanel Varnish Nginx Plugin ] - Enhance LiteSpeed and Apache Performance
    www.unixy.net - Los Angeles | Houston | Atlanta | Rotterdam
    Love to help pro bono (time permitting). joe > unixy.net
      0 Not allowed!

  32. #112
    Quote Originally Posted by UNIXy View Post
    I don't mind accepting the fact that Litespeed could possibly run faster (if it were to be proven).
    So, you do not mind taking a public shoot out between LiteSpeed and your Varnish plugin now.
    I have invited you before, unfortunately, you were not brave enough to answer it.
    http://www.webhostingtalk.com/showpo...&postcount=116

    The third party did apply a trial of your varnish plugin, our staff helped with LiteSpeed setup, that user run the "ab" tests himself and posted the results.

    We can use the VDS provided by Brent, or,I do not even mind doing it on your server/vps either, as long as third party can verify the setup.

    Quote Originally Posted by UNIXy View Post
    However, Litespeed Tech will never accept the fact that Varnish has performed faster. Someone in this forum ran benchmarks (a Litespeed Tech client mind you) and found Varnish to be 500 req/s faster than Litespeed Cache!
    Unfortunately, it is not easy to setup both correctly and have an apple to apple comparison and come up with the right conclusion. I hope a public shoot out will clear it up.

    BTW: that user is still using LiteSpeed.
    LiteSpeed Web Acceleration Platform by http://www.litespeedtech.com
    Apache drop-in replacement. Triple server capacity with 10X performance increases.
    Ultimate web serving platform for WordPress, Magento and other web applications.
    Turbo charging all WordPress sites hosted on your server with a single click!
      0 Not allowed!

  33. #113
    Join Date
    Jan 2011
    Location
    Ohio
    Posts
    467
    DO IT! DO A PUBLIC SHOOTOUT!
      0 Not allowed!

  34. #114
    Varnish is good, but Litespeed is better.
    Ryan C. | ComCraft Networks
    ComCraft Networks - Your Way To Host!
    Shared - VPS - Reseller - Dedicated - Semi-Dedicated - SEO - Minecraft - cPanel + Installatron and Softaculous - Colocation - Web Design - eCommerce - Live Support
    http://www.comcrafthosting.com/
      0 Not allowed!

  35. #115
    Join Date
    Feb 2008
    Location
    Houston, Texas, USA
    Posts
    2,955
    Quote Originally Posted by mistwang View Post
    So, you do not mind taking a public shoot out between LiteSpeed and your Varnish plugin now.
    Answer my question first and I'll be happy to help out with the benchmark. If you're not benchmarking the Unixy plugin on your forum, then why is it labeled as such? Change the title in your forum to reflect the fact that this is not unixy's plugin and we'll talk. Fair enough?

    Quote Originally Posted by mistwang View Post
    The third party did apply a trial of your varnish plugin, our staff helped with LiteSpeed setup, that user run the "ab" tests himself and posted the results.
    You know how I know this is not unixy's? Check the server header on Varnish. It seems like your "third party" has dropped the ball somewhere. Here's another tidbit for you (from Litespeed's forum):

    Server Software: LiteSpeed
    Document Length: 34243 bytes

    Server Software: Varnish
    Document Length: 39265 bytes

    Why are you making Varnish serve a larger document? That's about 5000 bytes more. As small as 5000Bytes is, it matters quite a lot in benchmarks. Before you blame it on Varnish, go check our benchmark results, both Varnish/Litespeed serve the same document length. Why so many red flags?

    Quote Originally Posted by mistwang View Post
    BTW: that user is still using LiteSpeed.
    Yes and I know exactly why that user is still using litespeed. Contrary to what you're insinuating, performance is not the reason ). I can't go over the details but the user started using the very early beta release of the plugin so little bugs are bound to crop up.
    UNIXy - Fully Managed Servers and Clusters - Established in 2006
    [ cPanel Varnish Nginx Plugin ] - Enhance LiteSpeed and Apache Performance
    www.unixy.net - Los Angeles | Houston | Atlanta | Rotterdam
    Love to help pro bono (time permitting). joe > unixy.net
      0 Not allowed!

  36. #116
    Join Date
    Jul 2005
    Location
    In the Internets
    Posts
    3,611
    Quote Originally Posted by UNIXy View Post
    Answer my question first and I'll be happy to help out with the benchmark. If you're not benchmarking the Unixy plugin on your forum, then why is it labeled as such? Change the title in your forum to reflect the fact that this is not unixy's plugin and we'll talk. Fair enough?



    You know how I know this is not unixy's? Check the server header on Varnish. It seems like your "third party" has dropped the ball somewhere. Here's another tidbit for you (from Litespeed's forum):

    Server Software: LiteSpeed
    Document Length: 34243 bytes

    Server Software: Varnish
    Document Length: 39265 bytes

    Why are you making Varnish serve a larger document? That's about 5000 bytes more. As small as 5000Bytes is, it matters quite a lot in benchmarks. Before you blame it on Varnish, go check our benchmark results, both Varnish/Litespeed serve the same document length. Why so many red flags?



    Yes and I know exactly why that user is still using litespeed. Contrary to what you're insinuating, performance is not the reason ). I can't go over the details but the user started using the very early beta release of the plugin so little bugs are bound to crop up.
    If you look at the post, you'll realize it wasn't him that did this test so you're ranting for no reason and looking stupid, imo.
      0 Not allowed!

  37. #117
    Join Date
    Nov 2005
    Location
    Portland, Oregon
    Posts
    1,080
    Quote Originally Posted by UNIXy View Post
    The title of the thread: LiteSpeed VS Unixy Varnish - The Truth!

    That's a lie!

    Whoever is running that show did not benchmark our plugin.

    They downloaded the default VCL from the Varnish repository, which everyone knows is suboptimal at best, and slapped the "Unixy" tag on it. Run the benchmark against the UNIXy plugin and get me involved. See I'm personally not tied to any one Web server or technology. Our clients run just all sorts of configurations from one VPS to clusters. Yes, some even run Litespeed!

    I don't mind accepting the fact that Litespeed could possibly run faster (if it were to be proven). However, Litespeed Tech will never accept the fact that Varnish has performed faster. Someone in this forum ran benchmarks (a Litespeed Tech client mind you) and found Varnish to be 500 req/s faster than Litespeed Cache!

    You'd think Litespeed Tech would stop misleading people with these lies. You know it's easy to tell from the benchmark output that this is not UNIXy's plugin, right? But you have the SEO upper arm right now so sadly folks will keep being misled.

    Regards
    We've used NGINX on all our servers for the past 6 months and are slowly migrating over to the UNIXY Varnish plugin. We operate our own hardware, so placing an additional 8GB of ram in our servers for caching is no problem. I've found that running the UNIXY Varnish plugin the load has always stayed under .5 (while memory is high, but that's not a problem). Previously, load spikes would incur anywhere from 3-10.

    Not to mention, it handles DDoS and any sort of floods much better than NGINX ever did for us. We use used Litespeed temporarily, poor results - it simply wasn't worth the money at all.
    » VPSFuze.com - Performance should be noticeable - VPS Hosting at its best.
    » HostingFuze.com - Affordable & Reliable Shared & Master Reseller hosting services
      0 Not allowed!

  38. #118
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    504
    So it sounds like UNIXy is onboard with an observed testing on our VDS.

    mistwang, you going to take the challenge?



    If mistwang says yes, I will need someone to provide websites that can be used for testing (no DNS changes needed, just a copy of some representative content of modern-day websites).
      0 Not allowed!

  39. #119
    Quote Originally Posted by brentpresley View Post
    So it sounds like UNIXy is onboard with an observed testing on our VDS.

    mistwang, you going to take the challenge?
    Yes, absolutely.
    LiteSpeed Web Acceleration Platform by http://www.litespeedtech.com
    Apache drop-in replacement. Triple server capacity with 10X performance increases.
    Ultimate web serving platform for WordPress, Magento and other web applications.
    Turbo charging all WordPress sites hosted on your server with a single click!
      0 Not allowed!

  40. #120
    Join Date
    Mar 2004
    Location
    Cheshire, UK & WA, USA
    Posts
    202
    Guys,

    I'll be setting the VDS up on one of our test grids. I think we need the community to decide the "rules" of the tests and we can then go ahead and setup everything that is needed etc.

    This experiment will be quite interesting, but I full expect Varnish to accelerate in some areas and varnish in others. It'll still come down to which products suit each company / server best.
      0 Not allowed!

Page 3 of 7 FirstFirst 123456 ... LastLast

Similar Threads

  1. LiteSpeed or Varnish, or what configuration
    By Serverfruit-Kris in forum Hosting Security and Technology
    Replies: 12
    Last Post: 03-13-2011, 11:13 AM
  2. Varnish VS Litespeed?
    By Dustin Cisneros in forum Hosting Software and Control Panels
    Replies: 33
    Last Post: 02-08-2011, 01:45 PM
  3. Replies: 0
    Last Post: 08-08-2010, 05:13 PM
  4. LiteSpeed Vs Apache?
    By leanfarrell in forum Hosting Security and Technology
    Replies: 11
    Last Post: 10-06-2009, 03:25 PM
  5. LiteSpeed Web Server vs. Apache 2x
    By Ivan23 in forum Hosting Software and Control Panels
    Replies: 3
    Last Post: 05-12-2005, 10:07 AM

Posting Permissions

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