Also, litespeed is practically a drop-in replacement for apache since it can obey httpd.conf and .htaccess files which no other webserver bothers to try to do. You can be converted to it in less than an hour (optionally along with apache in the background on another port). If you have a bunch of rewrite rules, other webservers are a complete workout to convert to. (They've also gone to great lengths to make it play nice with cpanel)
Re:Cherokee Webserver, "The Fastest", Really?
by StavrosK on Sunday March 13, 2011 @09:13AM
(#2319111)
Forgive my ignorance, but when does this actually matter? My application can't run anywhere near that speed, so time spent at the webserver level is less than 1% of the total response...
For static files or cached pages, I can maybe understand the fuss.
Re:Cherokee Webserver, "The Fastest", Really?
by adulau on Sunday March 13, 2011 @06:50AM
(#2318928)
"G-WAN is fully functional for an unlimited time at no cost for personal use, individual use, non-profit use, academic use, commercial and non-commercial use or any combination of these (no call-home, no spyware, no backdoor included)."
G-WAN is proprietary software, using "C" servlet for writing extension and is not very clear about the exact HTTP version supported.
On the other hand, Cherokee is free software (GNU General Public License), supports many handlers, includes many modules like reverse proxy and has a decent speed over Apache HTTPD.
Why would I pick the "The Fastest" when you have a full-featured, free, standard HTTP server that is already "Fast enough"?
Re:Cherokee Webserver, "The Fastest", Really?
by FooBarWidget on Sunday March 13, 2011 @06:33AM
(#2318906)
Nginx has an evented I/O architecture. libev/libevent implement evented I/O but evented architectures have got nothing to do with libev/libevent per se.
Re:Cherokee Webserver, "The Fastest", Really?
by piotrSikora on Sunday March 13, 2011 @06:43AM
(#2318916)
nginx uses native evented I/O provided by the operating system (epoll on Linux, kqueue on *BSD, /dev/poll and event ports on Solaris, input/output completion port on Windows).
libevent and libev provide abstraction layer for those mechanisms and unified interface for all supported operating systems.
Re:Cherokee Webserver, "The Fastest", Really?
by deno on Sunday March 13, 2011 @07:05AM
(#2318948)
> native evented I/O API provided by operating system
I know it's just semantics but I recently learned that some people actually believe that synchronous I/O API provided by OS is actually… well synchronous I/O. Obviously there cannot be any synchronous I/O going on in any modern operating system.
Because I believe litespeed is faster, though the free version has a simultaneous connection limit.
It's even faster than nginx in some cases.
http://blog.litespeedtech.com/2010/01/06/benchmark-compariso...
Also, litespeed is practically a drop-in replacement for apache since it can obey httpd.conf and .htaccess files which no other webserver bothers to try to do. You can be converted to it in less than an hour (optionally along with apache in the background on another port). If you have a bunch of rewrite rules, other webservers are a complete workout to convert to. (They've also gone to great lengths to make it play nice with cpanel)
For Rails work these days though, I'm using passenger with either nginx or apache.
http://gwan.ch/en_jihad.html http://gwan.ch/en_complexity.html
For static files or cached pages, I can maybe understand the fuss.
G-WAN is proprietary software, using "C" servlet for writing extension and is not very clear about the exact HTTP version supported.
On the other hand, Cherokee is free software (GNU General Public License), supports many handlers, includes many modules like reverse proxy and has a decent speed over Apache HTTPD.
Why would I pick the "The Fastest" when you have a full-featured, free, standard HTTP server that is already "Fast enough"?
libevent and libev provide abstraction layer for those mechanisms and unified interface for all supported operating systems.
I know it's just semantics but I recently learned that some people actually believe that synchronous I/O API provided by OS is actually… well synchronous I/O. Obviously there cannot be any synchronous I/O going on in any modern operating system.