Page 2 of 4 FirstFirst 1234 LastLast
Results 41 to 80 of 126
  1. #41
    I didn't said it cannot be done, just it is too much work to set it up.
    If you want to run nginx, you can run it without cpanel. I checked most popular CMS (drupal, joomla, wordpress) and for all of them rewrite rules are available for nginx.

    So in my case I have differentiate those 3 web servers,

    shared hosting, cpanel I use LiteSpeed
    big site on server nginx
    small sites that do not have much traffic apache.

  2. #42
    Join Date
    Oct 2002
    Location
    Tarragona, SPAIN
    Posts
    93
    Bono has explained pretty much why litespeed for us, shared hosting.

    In fact, we are currently using nginx now, but only for serving static content, dynamic content is passed to apache.

    We want to trial test litespeed soon to see if there is a big difference on dynamic content or if it is not worth it. NGINX for static files has supposed a big difference.

    Regards,
    www.FactoriaDigital.com - Application web hosting & specialized wordpress hosting, in spanish

  3. #43
    Join Date
    Oct 2002
    Location
    Tarragona, SPAIN
    Posts
    93
    Also, has someone tested litespeed (Dynamic) + nginx (static)? that's the config they use at wordpress.com, i think they also use varnish but not sure.

    Regards,
    www.FactoriaDigital.com - Application web hosting & specialized wordpress hosting, in spanish

  4. #44
    Join Date
    Oct 2002
    Location
    Tarragona, SPAIN
    Posts
    93
    I've seen this thread: http://www.webhostingtalk.com/showth...t=apache+nginx

    Here they say that a fine tuned apache 2.2 with worker mpm can have the same performance than litespeed for dynamic content.

    Opinions?
    www.FactoriaDigital.com - Application web hosting & specialized wordpress hosting, in spanish

  5. #45
    Join Date
    Mar 2008
    Posts
    1,717
    Quote Originally Posted by Factor View Post
    Here they say that a fine tuned apache 2.2 with worker mpm can have the same performance than litespeed for dynamic content.

    Opinions?
    It echoes our experience with it. Apache out of the box performs like garbage, but with some tuning it can be made to perform great (for but one example: a sick little Pentium4 went from server load ~4.5 to server load ~0.25 after only Apache tuning). If you're not interested in messing with tuning, pay money to Litespeed...
    I used to run the oldest commercial Mumble host.

  6. #46
    Join Date
    Aug 2002
    Location
    Seattle
    Posts
    5,512
    Quote Originally Posted by fwaggle View Post
    It echoes our experience with it. Apache out of the box performs like garbage, but with some tuning it can be made to perform great (for but one example: a sick little Pentium4 went from server load ~4.5 to server load ~0.25 after only Apache tuning). If you're not interested in messing with tuning, pay money to Litespeed...
    In the same spirit a tuned Litespeed will substantially out perform a tuned Apache.

  7. #47
    Join Date
    Feb 2006
    Location
    Buffalo NY
    Posts
    1,330
    Quote Originally Posted by IRCCo Jeff View Post
    In the same spirit a tuned Litespeed will substantially out perform a tuned Apache.
    Let's not get too crazy - any information / evidence to back this up? In what sense would it substantially outperform Apache? When serving static files? PHP files? Using what method (FCGI, Module, etc)?

    I'm a proponent of Litespeed, especially in a shared environment - that being said pissing matches / extraordinary claims require at least some benchmarking & evidence.
    Cody R.
    Hawk Host Inc. Proudly Serving websites since 2004.

  8. #48
    Join Date
    Mar 2008
    Posts
    1,717
    It might, but to me it's not worth the high cost of ownership on performance alone. Some of the other features are pretty nice, though.

    It just irks me how they compare litespeed to Apache's Prefork MPM, which everyone knows performs terribly. I don't consider myself an Apache tuning expert by any stretch of the imagination, and my Apache benchmarks pretty close to equal clear up to maxing out a 10mbps VPS I was running AB from, 800 requests concurrency 100. The numbers are very slightly in favor of litespeed of course, but not enough to warrant the boast of "2x as fast as Apache!"

    I still stand behind my original statement, that if you don't want to bother configuring Apache then Litespeed makes a great drop-in replacement, particularly for cPanel servers... the benchmarks just bug me though.
    I used to run the oldest commercial Mumble host.

  9. #49
    Quote Originally Posted by fwaggle View Post
    It might, but to me it's not worth the high cost of ownership on performance alone. Some of the other features are pretty nice, though.

    It just irks me how they compare litespeed to Apache's Prefork MPM, which everyone knows performs terribly. I don't consider myself an Apache tuning expert by any stretch of the imagination, and my Apache benchmarks pretty close to equal clear up to maxing out a 10mbps VPS I was running AB from, 800 requests concurrency 100. The numbers are very slightly in favor of litespeed of course, but not enough to warrant the boast of "2x as fast as Apache!"

    I still stand behind my original statement, that if you don't want to bother configuring Apache then Litespeed makes a great drop-in replacement, particularly for cPanel servers... the benchmarks just bug me though.
    Do you have any evidence of this? Apache can be tuned but I don't think it gets close to litespeed, there is a reason why people are willing to spend hundreds or even thousands for their license. By your comment, it sounds as they are just lazy and prefer to pay for a tuned web server out of the box, which doesn't make sense, since with the cost of litespeed you could hire the best apache expert and basically put your server in shape.

    I don't run litespeed, but im also wondered why nobody posted benchmarks except litespeed of course which is biased to their product. But you have a point here, if litespeed outperforms apache so match, it would be just plain silly that apache would decide not to implement similar features. I agree that Apache out of the box sucks, but im not sure how much you can push it to be close to litespeed without affecting stability. A fast server is worthless if it crashes all the time. Litespeed not only claims to outperform apache by 2 to 4 times but it also advertises its more secure and even can mitigate small dos attacks. I wonder what litespeed is really up to the task besides their nice admin tools. It would be nice to compare real live scenarios.

  10. #50
    Join Date
    Aug 2002
    Location
    Seattle
    Posts
    5,512
    Quote Originally Posted by CodyRo View Post
    Let's not get too crazy - any information / evidence to back this up? In what sense would it substantially outperform Apache? When serving static files? PHP files? Using what method (FCGI, Module, etc)?

    I'm a proponent of Litespeed, especially in a shared environment - that being said pissing matches / extraordinary claims require at least some benchmarking & evidence.
    I've not published a benchmark but my company is paying thousands each month for licensing on software I could be getting for free if it weren't actually performing as advertised.

    End game is that Litespeed is either a) really great or b) i'm really stupid.

  11. #51
    To start out, we switched to LiteSpeed after having muple server management companies look over a server that was appraching its "capacity" and tell us there was nothing we could do but upgrade the hardware to avoid any more problems. What did we do? Take advantage of the drop in replacement that is fully compatible with Apache (even reading Apache's config files!).

    To sum up our experience with Apache 2.2 Tweaked for performance/resources & LiteSpeed non-tweaked....

    LiteSpeed 4 enabled on the server, hardly any wait time, using 45% of the memory, and a load of two using a 1 core license on a 16 core machine.....

    Apache 2.2 tweaked (non-prefork...etc) RAM 100%, Load 76.0 within minutes and climbing by the minute... Apache finally becomes unstable and refuses to serve any longer... server was dead within 5 minutes of switching LiteSpeed out for Apache...
    We don't need benchmarks to prove LiteSpeed is better. We already know it.

    /End Thread?

  12. #52
    Join Date
    Jun 2006
    Location
    NYC
    Posts
    1,446
    Quote Originally Posted by IRCCo Jeff View Post
    I've not published a benchmark but my company is paying thousands each month for licensing on software I could be getting for free if it weren't actually performing as advertised.

    End game is that Litespeed is either a) really great or b) i'm really stupid.
    Not for us to judge but we've done benchmarks of our own and it was going to be really stupid of us to switch to Litespeed because (on average) Apache was serving our sites comparatively or faster.

    Litespeed has some advantages but it definitely has its faults. Somehow everyone overlooks the faults and talks about how great it is. It's great until they are sold or go under and then you're all left to try and keep your customers afloat when your licenses stop working.

    They state it VERY clear, "small company". Extremely small with 1 real developer. It's a dangerous tightrope any litespeed customer is walking. HyperVM all over again, imo. If they were a big business and had a real development staff then it wouldn't be such a concern. Although, I've been one of those people to place a ticket and the one guy who knows the answer is "out" for a few days. Scary people would put their business in the hands of a product lacking that much in terms of development.

    Not to say that Litespeed isn't an excellent product. Though, when compared to Apache, it isn't a feat of wonder Lighttpd and nginx easily compare and in terms of flexibility, Apache wins hands down.
    Last edited by FiberPeer; 10-06-2009 at 04:29 AM.
    FiberPeer.Com | | REAL DDoS Protection | Cloud Hosting | VPS | Dedicated Servers | High Bandwidth Hosting | 1Gbps-10Gbps Unmetered
    FiberPeer DDoS Mitigation | ethProxy Upgraded! | 14-Years Experience | Emergency 24/7 Support
    Visit us @ www.fiberpeer.com

  13. #53
    Join Date
    May 2008
    Location
    EU
    Posts
    94
    Based on info from LS forum, it appears that the latest version has some kind of bug which randomly lets people download PHP files instead of executing them:
    http://www.litespeedtech.com/support...?t=3101&page=5

    And there hasn't been any releases after this bug report, not sure how serious is this, but I wouldn't want my php files to be downloaded.

    Btw twitter used Litespeed at some point, but if you check what server they are running now, it says "Hi".

  14. #54
    Join Date
    Oct 2007
    Posts
    4,016
    Quote Originally Posted by Sergej View Post
    Based on info from LS forum, it appears that the latest version has some kind of bug which randomly lets people download PHP files instead of executing them:
    http://www.litespeedtech.com/support...?t=3101&page=5

    And there hasn't been any releases after this bug report, not sure how serious is this, but I wouldn't want my php files to be downloaded.

    Btw twitter used Litespeed at some point, but if you check what server they are running now, it says "Hi".
    This is not true. The said thread is discussing about version 4.1RC1.
    "RC" means "Release Candidate", which is not a stable version and should not be used in a production environment.

    The current latest stable version is 4.0.10.

  15. #55
    Join Date
    Jan 2004
    Location
    South East U.K.
    Posts
    1,295
    Quote Originally Posted by JLHC View Post
    This is not true. The said thread is discussing about version 4.1RC1.
    "RC" means "Release Candidate", which is not a stable version and should not be used in a production environment.

    The current latest stable version is 4.0.10.
    True, but if you read on someone states they've seen this behaviour on 4.0.10

  16. #56
    Join Date
    Oct 2007
    Posts
    4,016
    Quote Originally Posted by RandomLittleHost View Post
    True, but if you read on someone states they've seen this behaviour on 4.0.10
    Well I have not tested 4.0.10 fully as I prefer to use the most stable version of 4.0.6. Besides that, 4.0.11 will be released soon to fix this issue.

  17. #57
    Quote Originally Posted by AquariusADMIN View Post
    To start out, we switched to LiteSpeed after having muple server management companies look over a server that was appraching its "capacity" and tell us there was nothing we could do but upgrade the hardware to avoid any more problems. What did we do? Take advantage of the drop in replacement that is fully compatible with Apache (even reading Apache's config files!).

    To sum up our experience with Apache 2.2 Tweaked for performance/resources & LiteSpeed non-tweaked....

    LiteSpeed 4 enabled on the server, hardly any wait time, using 45% of the memory, and a load of two using a 1 core license on a 16 core machine.....

    Apache 2.2 tweaked (non-prefork...etc) RAM 100%, Load 76.0 within minutes and climbing by the minute... Apache finally becomes unstable and refuses to serve any longer... server was dead within 5 minutes of switching LiteSpeed out for Apache...
    We don't need benchmarks to prove LiteSpeed is better. We already know it.

    /End Thread?
    I wonder what do you mean with tweaked?
    There must have being something wrong on your apache config, what you basically say is litespeed performs 38 times better. Thats to good to be truth. I really wish this is truth, because then we are all stupids running apache. 90% of webhosting companies instead of deploying more servers would just implement litespeed.

    Some people share your opinion, and then others say apache 2 with worker if tuned runs with the same performance.

    I wonder then if litespeed can be also tuned, as you said LiteSpeed non-tweaked.

    Is this on a php dynamic environment? So you run it on only 1 core without any issues even when its on a 16 core server?
    Last edited by PYDOT; 10-06-2009 at 02:23 PM.

  18. #58
    Quote Originally Posted by ServerOrigin View Post
    Not for us to judge but we've done benchmarks of our own and it was going to be really stupid of us to switch to Litespeed because (on average) Apache was serving our sites comparatively or faster.

    Litespeed has some advantages but it definitely has its faults. Somehow everyone overlooks the faults and talks about how great it is. It's great until they are sold or go under and then you're all left to try and keep your customers afloat when your licenses stop working.

    They state it VERY clear, "small company". Extremely small with 1 real developer. It's a dangerous tightrope any litespeed customer is walking. HyperVM all over again, imo. If they were a big business and had a real development staff then it wouldn't be such a concern. Although, I've been one of those people to place a ticket and the one guy who knows the answer is "out" for a few days. Scary people would put their business in the hands of a product lacking that much in terms of development.

    Not to say that Litespeed isn't an excellent product. Though, when compared to Apache, it isn't a feat of wonder Lighttpd and nginx easily compare and in terms of flexibility, Apache wins hands down.
    Everybody starts small some day, that doesn't mean they don't deserve a chance. But I agree, if thats is the case I would not use it for high sensitive services.

  19. #59
    Join Date
    Aug 2004
    Location
    Canada
    Posts
    3,571
    Quote Originally Posted by PYDOT View Post
    Is this on a php dynamic environment? So you run it on only 1 core without any issues even when its on a 16 core server?
    The number of core licenses is how many processes of lshttpd there will be. So if most of the CPU is dynamic content then one core license is fine. If you have a lot of static content you may need more lshttpd processes available.

    The model it uses is similar to this experimental mpm: http://httpd.apache.org/docs/2.2/mod/event.html. Same idea as nginx and lighttpd as well.

    If people say their Apache performs better than litespeed I assume they can make it perform better than nginx and lighttpd as well?
    Tony B. - Chief Executive Officer
    Hawk Host Inc. Proudly serving websites since 2004
    Quality Shared and VPS Hosting
    PHP 5.3.x & PHP 5.4.x & PHP 5.5.X & PHP 5.6.X & PHP 7.0.X Support!

  20. #60
    Join Date
    Dec 2005
    Posts
    1,270
    Who is Litespeed to tell me that porn "deteriorates human moral values". I wish them nothing other than failure. I hate it when people push there moral feeling on others. Companies like these don't go every far.

  21. #61
    Join Date
    Mar 2009
    Posts
    649
    Quote Originally Posted by mohamoud View Post
    Who is Litespeed to tell me that porn "deteriorates human moral values". I wish them nothing other than failure. I hate it when people push there moral feeling on others. Companies like these don't go every far.
    Their assumption that they have the right to determine "morality" is the reason I will never purchase Litespeed, regardless of performance. That said, it doesn't have much to do with performance. You could, however, argue that a full benchmark of Litespeed is impossible, because one is disallowed from benchmarking its performance serving the majority of web traffic.
    [GB ≠ GiB] [MB ≠ MiB] [kB ≠ kiB] [1000 ≠ 1024] [Giga ≠ gram] [Mega ≠ milli] [Kelvin ≠ kilo] [Byte ≠ bit]
    There is no millibit. There is no gram-bit. There is no Kelvin-Byte.

  22. #62
    Quote Originally Posted by Sergej View Post
    Based on info from LS forum, it appears that the latest version has some kind of bug which randomly lets people download PHP files instead of executing them:
    http://www.litespeedtech.com/support...?t=3101&page=5

    And there hasn't been any releases after this bug report, not sure how serious is this, but I wouldn't want my php files to be downloaded.

    Btw twitter used Litespeed at some point, but if you check what server they are running now, it says "Hi".
    IN LSWS switching between versions is matter of seconds, so you can have few versions precompiled and just switch between them without any downtime.
    I'm currently using 4.0.6 version and I can say I'm happy with it, under DDOS my sites stay responsive which I couldn't say about Apache.

  23. #63
    Join Date
    Aug 2002
    Location
    Seattle
    Posts
    5,512
    Quote Originally Posted by Bono_ View Post
    under DDOS my sites stay responsive which I couldn't say about Apache.
    Bingo. My staff is skilled in optimizing both Apache and LSWS. We've been able to handle 8000 requests per second on a LSWS server that was crawling when running Apache with < 200 requests.

  24. #64
    Join Date
    Mar 2009
    Posts
    649
    IRCCo Jeff: Less extraordinary claims, and more evidence, please.
    [GB ≠ GiB] [MB ≠ MiB] [kB ≠ kiB] [1000 ≠ 1024] [Giga ≠ gram] [Mega ≠ milli] [Kelvin ≠ kilo] [Byte ≠ bit]
    There is no millibit. There is no gram-bit. There is no Kelvin-Byte.

  25. #65
    Join Date
    Aug 2002
    Location
    Seattle
    Posts
    5,512
    Quote Originally Posted by petteyg359 View Post
    IRCCo Jeff: Less extraordinary claims, and more evidence, please.
    I'm not motivated enough to try to convince you to switch. If you don't want to believe it just carry on with Apache.

  26. #66
    Join Date
    Oct 2008
    Posts
    2,249
    Quote Originally Posted by IRCCo Jeff View Post
    Bingo. My staff is skilled in optimizing both Apache and LSWS. We've been able to handle 8000 requests per second on a LSWS server that was crawling when running Apache with < 200 requests.
    I have crawled on apache with that many request and I have been able to handle 2 thousand with litespeed before it crawled... Although I was running a low end dedicated.
    Leader of the new anti sig spamming club.

  27. #67
    Join Date
    Mar 2009
    Posts
    649
    Quote Originally Posted by IRCCo Jeff View Post
    I'm not motivated enough to try to convince you to switch. If you don't want to believe it just carry on with Apache.
    Okay, I get it. We're supposed to take it on "faith". No wonder the license disallows porn. It's a religious web server.
    [GB ≠ GiB] [MB ≠ MiB] [kB ≠ kiB] [1000 ≠ 1024] [Giga ≠ gram] [Mega ≠ milli] [Kelvin ≠ kilo] [Byte ≠ bit]
    There is no millibit. There is no gram-bit. There is no Kelvin-Byte.

  28. #68
    Join Date
    Aug 2002
    Location
    Seattle
    Posts
    5,512
    Quote Originally Posted by darkeden View Post
    I have crawled on apache with that many request and I have been able to handle 2 thousand with litespeed before it crawled... Although I was running a low end dedicated.
    2000 is the hard limit if you don't tune it.

  29. #69
    Join Date
    Jan 2005
    Posts
    2,175
    Quote Originally Posted by IRCCo Jeff View Post
    Bingo. My staff is skilled in optimizing both Apache and LSWS. We've been able to handle 8000 requests per second on a LSWS server that was crawling when running Apache with < 200 requests.
    Quote Originally Posted by petteyg359 View Post
    IRCCo Jeff: Less extraordinary claims, and more evidence, please.
    http://www.litespeedtech.com/support...9&postcount=16

    I suggest you read the entire thread.

  30. #70
    Join Date
    Mar 2009
    Posts
    649
    That thread is evidence of nothing. Show me a benchmark.
    [GB ≠ GiB] [MB ≠ MiB] [kB ≠ kiB] [1000 ≠ 1024] [Giga ≠ gram] [Mega ≠ milli] [Kelvin ≠ kilo] [Byte ≠ bit]
    There is no millibit. There is no gram-bit. There is no Kelvin-Byte.

  31. #71
    Join Date
    Jan 2005
    Posts
    2,175
    I think real world examples are more convincing than benchmarks. Are you saying that those people lied?

  32. #72
    Join Date
    Mar 2009
    Posts
    649
    I an example of somebody thinking that switching to another HTTP server is the answer to a DoS. I don't see any information on what he's actually serving. Plain HTML? PHP with LS SAPI? PHP through FastCGI with PHP-FPM, spawn-fcgi, or daemonize? Ruby through FastCGI? CGI through FastCGI? WSGI? SCGI? Was the DoS still occurring at the exact same level once LS was installed? Were any iptables or firewall stuff changed?

    In other words, useless for actually comparing performance.
    [GB ≠ GiB] [MB ≠ MiB] [kB ≠ kiB] [1000 ≠ 1024] [Giga ≠ gram] [Mega ≠ milli] [Kelvin ≠ kilo] [Byte ≠ bit]
    There is no millibit. There is no gram-bit. There is no Kelvin-Byte.

  33. #73
    Quote Originally Posted by petteyg359 View Post
    Okay, I get it. We're supposed to take it on "faith". No wonder the license disallows porn. It's a religious web server.
    Please download LSWS trial which lasts for 14 days, tweak apache and try to beat LSWS in performance.
    Post your results and prove that thousands of users who pay for LSWS are actually idiots who cannot optimize Apache.

  34. #74
    Join Date
    May 2008
    Location
    EU
    Posts
    94
    Yeah the trial license is 2 core enterprise version.

  35. #75
    Join Date
    Mar 2008
    Posts
    1,717
    Quote Originally Posted by Bono_ View Post
    Please download LSWS trial which lasts for 14 days, tweak apache and try to beat LSWS in performance.
    Post your results and prove that thousands of users who pay for LSWS are actually idiots who cannot optimize Apache.
    I'm not saying they're idiots by any stretch of the imagination, my only gripe about performance is that LiteSpeed base their benchmarks on the Prefork MPM which only really describes it's effectiveness to the exact group of people you just described. If LSWS works out superior for you, that's awesome - but it's not for everyone.

    Who knows, if we start selling a bunch of webhost packages (web hosting isn't our primary business market at the moment) I might wind up needing to switch. However for our current use, Apache's working reasonably well for us given that most of what it's serving is static content. Comparing Apache/Litespeed to lighttpd/nginx is apples-to-oranges, you're comparing full-featured serving platforms to very nice pieces of software that are unfortunately lacking in some features.

    Find below benchmarks that I messed around with, I do understand they're not definitive by any stretch of the imagination. I made no attempts to tune LiteSpeed at all, but then again our Apache 2.2 is hardly tuned either (more or less just Worker MPM and the other usual suspects). I also fully realize it's quite low bandwidth and low concurrency compared to where LSWS really shines, but this was the best I could manage without having the VPS I was running the benchmarks from run out of steam (numbers got higher on both, without the load going up on the server significantly).

    It shows that for my purposes (obviously not the same as everyone's), LSWS is not the obvious improvement everyone seems to make out it is:

    Code:
    Server Software:        Apache/2.2.11
    Server Port:            80
    
    Document Path:          /index.html
    Document Length:        834 bytes
    
    Concurrency Level:      100
    Time taken for tests:   0.859599 seconds
    Complete requests:      800
    Failed requests:        0
    Write errors:           0
    Total transferred:      931563 bytes
    HTML transferred:       668034 bytes
    Requests per second:    930.67 [#/sec] (mean)
    Time per request:       107.450 [ms] (mean)
    Time per request:       1.074 [ms] (mean, across all concurrent requests)
    Transfer rate:          1057.47 [Kbytes/sec] received
    
    Connection Times (ms)
                  min  mean[+/-sd] median   max
    Connect:       45   46   0.2     46      47
    Processing:    46   49   5.5     48      70
    Waiting:       46   49   5.3     47      69
    Total:         92   95   5.5     94     116
    
    Percentage of the requests served within a certain time (ms)
      50%     94
      66%     94
      75%     95
      80%     95
      90%    103
      95%    112
      98%    114
      99%    115
     100%    116 (longest request)
    
    
    Server Software:        LiteSpeed
    Server Port:            8088
    
    Document Path:          /index.html
    Document Length:        834 bytes
    
    Concurrency Level:      100
    Time taken for tests:   0.844408 seconds
    Complete requests:      800
    Failed requests:        0
    Write errors:           0
    Total transferred:      857600 bytes
    HTML transferred:       667200 bytes
    Requests per second:    947.41 [#/sec] (mean)
    Time per request:       105.551 [ms] (mean)
    Time per request:       1.056 [ms] (mean, across all concurrent requests)
    Transfer rate:          991.23 [Kbytes/sec] received
    
    Connection Times (ms)
                  min  mean[+/-sd] median   max
    Connect:       45   46   0.2     46      47
    Processing:    46   47   1.7     47      54
    Waiting:       46   47   1.4     47      54
    Total:         92   93   1.7     94     100
    
    Percentage of the requests served within a certain time (ms)
      50%     94
      66%     94
      75%     94
      80%     94
      90%     95
      95%     98
      98%     99
      99%    100
     100%    100 (longest request)
    I used to run the oldest commercial Mumble host.

  36. #76
    Join Date
    Aug 2004
    Location
    Canada
    Posts
    3,571
    You know what since this topic is going nowhere and no one really wants to benchmark we (Hawk Host) will do it. We obviously have some bias though but we don't have a stake on the actual web servers.

    Just so we avoid all the you did not test x configuration or you did not configure it properly. Feel free to post any configuration file with tweaks you want to see things tested. Also depending on the web server any mpm's.

    We'll test apache, nginx, lighttpd and litespeed. Any alternative reasonable mpm's we'll run those as well. For the PHP tests the various ways to run it. If you have any other random web servers we could probably run those as well.

    As far as the test itself any concurrency level or whatever you want just say we'll do it.

    The test machine is a modest Opteron 170 with 2GB of ram. It's not some massive Dual 5550 with 24gb of ram but it should suffice for testing.
    Tony B. - Chief Executive Officer
    Hawk Host Inc. Proudly serving websites since 2004
    Quality Shared and VPS Hosting
    PHP 5.3.x & PHP 5.4.x & PHP 5.5.X & PHP 5.6.X & PHP 7.0.X Support!

  37. #77
    Join Date
    Mar 2009
    Posts
    649
    Unless somebody can make Apache, nginx, lighttpd, etc. talk to LS SAPI, benchmark should be done with FastCGI PHP (spawn-fcgi or PHP-FPM).
    [GB ≠ GiB] [MB ≠ MiB] [kB ≠ kiB] [1000 ≠ 1024] [Giga ≠ gram] [Mega ≠ milli] [Kelvin ≠ kilo] [Byte ≠ bit]
    There is no millibit. There is no gram-bit. There is no Kelvin-Byte.

  38. #78
    Join Date
    Mar 2008
    Posts
    1,717
    Tony: I would be really curious to see the results of that, and would genuinely appreciate you doing it. I would be very interested to see the turning point at which LSWS becomes a true advantage - even though I'm an Apache fanboy I do acknowledge that until Event MPM goes production, LSWS will outrun Apache eventually. I don't see the point in testing nginx/lighttpd, because as I said earlier they're not full-featured webservers... they won't function as a drop-in replacement for Apache, whereas LSWS will. But hey if you want to spend the time to do it, that'd be nice.

    I would benchmark it myself, but I have two problems: one, I don't have a better host to run the benchmark from and two, I don't have a true test server... I was benchmarking our production server so I didn't want to belt it too hard.

    I don't particularly care too much as to the tuning of Apache - as long as it's not configured to run on Prefork. Up to a concurrency level of 100~200, Apache should mostly keep up with LSWS... if it fails to do that for you it's a pretty good indication you're probably doing something wrong.

    Assuming your Apache configuration doesn't rely on mod_ssl, can you benchmark MPM Event too? That would be extremely interesting to me.
    I used to run the oldest commercial Mumble host.

  39. #79
    Join Date
    Aug 2004
    Location
    Canada
    Posts
    3,571
    Quote Originally Posted by petteyg359 View Post
    Unless somebody can make Apache, nginx, lighttpd, etc. talk to LS SAPI, benchmark should be done with FastCGI PHP (spawn-fcgi or PHP-FPM).
    The PHP comparison will be their SAPI against others. They say it's faster than mod_php and fastcgi in apache. It won't give you a conclusive answer though if lsws is faster than the web server at static it has an advantage in php to begin with. But seeing as people are saying they can make Apache + FastCGI faster than LiteSpeed's version I think it has to be benchmarked.


    Quote Originally Posted by fwaggle View Post
    Tony: I would be really curious to see the results of that, and would genuinely appreciate you doing it. I would be very interested to see the turning point at which LSWS becomes a true advantage - even though I'm an Apache fanboy I do acknowledge that until Event MPM goes production, LSWS will outrun Apache eventually. I don't see the point in testing nginx/lighttpd, because as I said earlier they're not full-featured webservers... they won't function as a drop-in replacement for Apache, whereas LSWS will. But hey if you want to spend the time to do it, that'd be nice.

    I would benchmark it myself, but I have two problems: one, I don't have a better host to run the benchmark from and two, I don't have a true test server... I was benchmarking our production server so I didn't want to belt it too hard.

    I don't particularly care too much as to the tuning of Apache - as long as it's not configured to run on Prefork. Up to a concurrency level of 100~200, Apache should mostly keep up with LSWS... if it fails to do that for you it's a pretty good indication you're probably doing something wrong.

    Assuming your Apache configuration doesn't rely on mod_ssl, can you benchmark MPM Event too? That would be extremely interesting to me.
    The other web servers are just there for comparison sake. The old litespeed benchmarks claimed it was faster than some of those as well.

    I think you're exactly right on what is going to happen. They'll be pretty much the same although litespeed will use less resources on concurrency level of 100-200. It will start to shine when Apache starts hitting the limits of the server.

    As far as the mpm event that was a given. It obviously is not production ready but it'll give a much better fight or maybe even faster than litespeed.
    Tony B. - Chief Executive Officer
    Hawk Host Inc. Proudly serving websites since 2004
    Quality Shared and VPS Hosting
    PHP 5.3.x & PHP 5.4.x & PHP 5.5.X & PHP 5.6.X & PHP 7.0.X Support!

  40. #80
    May be i am not the average users but, in my experience, the real bottleneck is the bandwidth and the database (cpu and memory).

    For a entry level, i would put a extra 1gb (considering that the base is usually 512-1gb) instead of spend for a litespeed license.

Page 2 of 4 FirstFirst 1234 LastLast

Similar Threads

  1. High Performance Litespeed Powered Hosting Resellers
    By jon-f in forum Reseller Hosting Offers
    Replies: 0
    Last Post: 07-02-2009, 12:26 PM
  2. Secureservertech - High Performance Litespeed Hosting
    By jon-f in forum Shared Hosting Offers
    Replies: 0
    Last Post: 07-01-2009, 05:20 PM
  3. High Performance Litespeed Hosting Solutions
    By jon-f in forum Shared Hosting Offers
    Replies: 0
    Last Post: 08-08-2008, 04:29 AM
  4. High Performance Litespeed Powered Hosting
    By jon-f in forum Shared Hosting Offers
    Replies: 0
    Last Post: 05-11-2008, 06:05 PM
  5. High Performance Cpanel/WHM Resellers | Litespeed Powered
    By jon-f in forum Reseller Hosting Offers
    Replies: 0
    Last Post: 10-27-2007, 01:45 PM

Posting Permissions

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