XenServ.com Litespeed Usage

cmanns

Well-Known Member
#1
When I ordered my license I mentioned to LiteSpeed we'd be coming up with some munin graphs of differences sadly didn't have any apache/cpu cacti graphs since cacti's much prettier we hope to get that done and see what SNMP functions litespeed has :)

So as my promise heres our story.

Now back in Feb I believe XenServ tried a trial of LiteSpeed, performed roughly same (per page load) we mostly host dynamic web sites like blogs and forums with XCache and Memcache. The most favorite part was no more 500 Errors randomly from apache 2.2 mod_fcgid. We did however like the stability, memory management and WHM/CPanel interface was quite clean functioning.
We decided the cost was not worth vs our income on the server, shortly after one client started experiencing extreme bursting where people would refresh, etc and it'd just generally load up the system from a not so large site. Another site a Wordpress blog generating well over 20-30EA Req/sec from Litespeeds admin view.

The issue boiled down to apache not liking backlogged external app requests I believe, or just not enough ram in general for apache we're quite happy with LiteSpeeds improvements for dynamic content delivery.

Our hardware is a dual socket xeon (4 core) system w/one avail cpu socket. 4GB dual channel ddr ecc registered and SATA2 drives on gigE uplinks. We utilize a 2 core license from LiteSpeed as it was told that there is FastCGI optimizations.

I do understand I see these graphs everyday and know more of the trend and some of our clients, I apologize for that. Also munin isn't very detailed however you should notice connections is smooth, ram is level and btw that was without PHP caching, now we run php caching and it's still pretty smooth, however it NEVER jumps up.



Another sample image of apache, connections just skyrocket along with maxclients, this issue is gone since litespeed, we utilize memcache/remote mysql so of course connections goes higher naturally :rolleyes:

http://uploadpla.net/files/6310_qolci/harbl.xenserv.com-netstat-day.png

Towards the end I was utilizing Apache 2.2 MPM_Event as another note, I spent well over a month trying to fix, the awful mod_fcgid only does 1 connection per process? So it always wants to spawn unless you allow lots of processes, and with XCache or any sort of caching like APC - bye bye ram, we'd need at least 6GB to run apache + mod_fcigd & Xcache, which may barely host our load compared to LiteSpeed 2 core license at 48/month plows through this load w/out issue, and helped us gain back ram to grow till we add more memory (Oh did I mention we were able to load XCache further more lowering load?)

I hope you guys enjoyed my case study, I'm very unbiased coming from a dedicated Lighttpd user till we migrated to Cpanel for shared, I must say your product is a very scalable solution esp for high dynamic load cases.

- William C. Manns
XenServ Hosting
 
Top