XenForo optimised VPS environment.

[Deleted member 745]

It should be even faster now Shamil I forgot to enable APC in the config file...http://xftest.radonsystems.net/threads/apc-now-enabled.8/#post-48

Holy mother its fast as hell now...

LiteSpeed (from today) --> Current load: 0.00, Max: 0.6, Avg: 0.03

LiteSpeed + APC rule

 

[Shamil]

It should be even faster now Shamil I forgot to enable APC in the config file...http://xftest.radonsystems.net/threads/apc-now-enabled.8/#post-48

Holy mother its fast as hell now...

LiteSpeed (from today) --> Current load: 0.00, Max: 0.6, Avg: 0.03

LiteSpeed + APC rule

I was going to open a support ticket for on-site optimisations actually. I'll still open it - it doesn't finish there

 

[Deleted member 745]

It's doing the same with me again where it wont let me upload images....

 

[SneakyDave]

I have a VPS with nginx, php fastcgi, and APC enabled, I might never go back to Apache. I'm also looking at cherokee and lightspeed. I think there are some specific xF APC parameters that can be played with(?), but I haven't tried them.

I have a random problem once in a while where one of the fastcgi processes takes up 100% CPU, and I have to restart all the fastcgis to resolve it. I have no idea what's causing it. Nothing in the php error logs, nothing in the nginx logs. I don't know how to tie the process number to a particular site or php file that is causing the problem.

Supposedly, it could be a permissions problem, but I don't see any errors anywhere indicating that. Does the fastcgi processes themselves have a stderr somewhere?

 

[Shamil]

I have a VPS with nginx, php fastcgi, and APC enabled, I might never go back to Apache. I'm also looking at cherokee and lightspeed. I think there are some specific xF APC parameters that can be played with(?), but I haven't tried them.

I have a random problem once in a while where one of the fastcgi processes takes up 100% CPU, and I have to restart all the fastcgis to resolve it. I have no idea what's causing it. Nothing in the php error logs, nothing in the nginx logs. I don't know how to tie the process number to a particular site or php file that is causing the problem.

Supposedly, it could be a permissions problem, but I don't see any errors anywhere indicating that. Does the fastcgi processes themselves have a stderr somewhere?

In our initial testing, LiteSpeed was better, overall, than Nginx. For static files, of course Nginx was superior.

 

[SneakyDave]

Yes, that's what I've seen from reports also. Does LiteSpeed have a php load module, or do you use PHP like nginx? Cherokee seemed to be getting a lot of good reviews, then LiteSpeed seemed to come out of nowhere with improvements.

 

[Deebs]

In our initial testing, LiteSpeed was better, overall, than Nginx. For static files, of course Nginx was superior.

Can you provide further details of your testing? What was the load? What about when LiteSpeed and nginx are serving 1000 concurrent requests each, what is the load then? Config files? Running php-fpm with nginx or fastcgi? Personally I cannot believe that a drop in replacement for Apache can be more efficient than nginx (which is so lightweight in comparison).

Happy to be proved wrong.

 

[Shamil]

Yes, that's what I've seen from reports also. Does LiteSpeed have a php load module, or do you use PHP like nginx? Cherokee seemed to be getting a lot of good reviews, then LiteSpeed seemed to come out of nowhere with improvements.

LiteSpeed has it's own implementation of mod_php, SAPI/PHP: http://www.litespeedtech.com/php-litespeed-sapi.html

Can you provide further details of your testing? What was the load? What about when LiteSpeed and nginx are serving 1000 concurrent requests each, what is the load then? Config files? Running php-fpm with nginx or fastcgi? Personally I cannot believe that a drop in replacement for Apache can be more efficient than nginx (which is so lightweight in comparison).

Happy to be proved wrong.

Hi Deebs, I can provide more details:

We looked at static file requests, using ApacheBench, and other tools. The requests per section for Nginx, where static files were concerned was higher than that for LiteSpeed, however, where the page was generated using PHP, LiteSpeed provided better results, sans-optimisation. With Nginx, PHP was running via php-fpm, though we hadn't tested FCGI yet.

 

[Deebs]

LiteSpeed has it's own implementation of mod_php, SAPI/PHP: http://www.litespeedtech.com/php-litespeed-sapi.html



Hi Deebs, I can provide more details:

We looked at static file requests, using ApacheBench, and other tools. The requests per section for Nginx, where static files were concerned was higher than that for LiteSpeed, however, where the page was generated using PHP, LiteSpeed provided better results, sans-optimisation. With Nginx, PHP was running via php-fpm, though we hadn't tested FCGI yet.

I would not waste time on php-fcgi, it is slower than PHP-FPM. Can you provide more details? ie. the output of your ApacheBench benchmarks and whatever other tools you used. Also, configs used would be great to see as well....

 

[Shamil]

I would not waste time on php-fcgi, it is slower than PHP-FPM. Can you provide more details? ie. the output of your ApacheBench benchmarks and whatever other tools you used. Also, configs used would be great to see as well....

I used standard configuration, out of the box, no optimisations to either platform at the time. I'll get the results - [I didn't perform the tests myself - a member of my support team did].

 

[Deebs]

I used standard configuration, out of the box, no optimisations to either platform at the time. I'll get the results - [I didn't perform the tests myself - a member of my support team did].

Thanks, that would be interesting reading material to say the least. Btw, where are you hosting in London? Telehouse or Sov? Most people are with one of those two....

 

[Shamil]

Thanks, that would be interesting reading material to say the least. Btw, where are you hosting in London? Telehouse or Sov? Most people are with one of those two....

The current server setup is in Scranton. We'll be putting a few servers in Maidenhead/Kent or Bluesquare/London.

 

[Deebs]

Ah the outer edge data centres. Forgive me for saying but I have always been wary of them as they do not have LINX connectivity and that is important for me. Out of interest, who are you using for transit at those sites?

 

[Shamil]

At Maidenhead, it'll be TiNet and NTT through to Tiscali. (Bluesquare)
At Bluesquare/London, actually, it won't be Bluesquare, it'll be TelecityGroup. That'll use either Level 3, LINX or AT&T. (haven't made a decision on which one).

 

[Deebs]

Tiscali suck. Their transit and local loop bandwidth in the UK is "not best".

 

[Shamil]

Tiscali suck. Their transit and local loop bandwidth in the UK is "not best".

It's not the best, however, it's stable with an interlink. The Tiscali network probably won't be given priority - the datacenter are considering on stepping them down.

 

[Deebs]

That would be wise... Also, just to note, if you or your transit provider are not a member of LINX then you will not gain anything from direct peering within their infrastructure.

 

[Shamil]

That would be wise... Also, just to note, if you or your transit provider are not a member of LINX then you will not gain anything from direct peering within their infrastructure.

BlueSquare definitely are (check BlueConnex).

 

[Sir nick]

I can't edit my first post, however, I should mention that I am primarily aiming to offer XenForo optimised environments from $20 per month. (cPanel additionally for $15). This will include LiteSpeed Web Server. Alternatively, for $15 per month, I will be able to offer Apache + Varnish.

So far, the LiteSeed (+ internal cache) is working very well.

Varnish is a very well built product I look forward to offering it to my clients to

 

[DregondRahl]

I prefer Nginx for easier rewrites and PHP-FPM is quite usful for large sites, as the PHP processes are dynamically created. Lighty runs php-cgi as fastCGI and its static. To be honest Lighty and Nginx are very close memory and CPU wise, but I prefer Nginx+PHP-FPM. Lighty gave me quite a list of trouble when it came to rewrite rules with Joomla, had to use LUA for it. Also Nginx has a native module for Memcached which is quite helpful. Static files are good, but reading from memory is even faster given the situation. I think Nginx is good with long-polling and can run as a push-server.


I will try Lighty's SAPI config sometime soon and see how it can perform. As for PHP Optimizers xCache is superior to APC for memory. ON a test server with XenForo it gave around 500KB difference per page load. Seems small but its actually quite a bit.