Benchmark

[andy]

I'm a bit surprised by the performance advantage Litespeed/FCGI/PHP has over Apache/mod_php as shown in your benchmark. After all, in Apache/mod_php the PHP engine is in-process, while in Litespeed/FCGI/PHP, there's inter-process communication costs between Apache & FCGI. I'd think Apache/mod_php will always be faster for that reason. Do you know what's behind the performance advantage?

Also do you have data for higher concurrency, maybe up to 100 concurrent clients? A comparison with Zeus/FCGI would be nice too.

Now that I think about it more, one way for Litespeed/FCGI/PHP to beat Apache/mod_php would be if you implement your FCGI as a single process multiplexing multiple concurrent FCGI requests from Litespeed, kinda like how Litespeed multiplex concurrent http requests. I don't think this is the case from reading the doc, but it'd be very cool. Do you have any plan to do this?

About SSL performance - why is Litespeed so much better than Apache in the benchmark?

 

[mistwang]

Theoretically, in-process communication should be faster than inter-process communication. I think the reason why litespeed is faster is because it is highly optimized, and apache's inter-module communication is not that efficient.

The result is similar with higher concurrency. LiteSpeed performance should be in the same neighborhood with Zeus/FCGI.

We do not implement FCGI multiplexing over one connection yet and PHP FCGI does not support it either. Yes, we plan to support it in the future.

I think Apache mod_ssl is not that efficient, for example, it has to store shared session cache on disk.

Most of our test was performed on low-end hardware, the performance difference will be even bigger on better hardware and our upcoming Enterprise edition will be even better.

You are welcome to benchmark it and post your result here.