Friday, 2 May 2014

Web Framework benchmark

Some interesting figures. Certainly not the best real world scenario with the repeated "hello world" but that give some insight as to what each can give us in term of performance.

I particular like the comparison between the Java frameworks, netty only adds a slight overhead on top of servlet, which get pretty hammered by the overhead vertx adds onto it. While Spring, on its side, ruins the very efficient pure servlet implementation.

Also happy to see that node.js is not that fast compared to good old JVM based languages.

RPS is request per seconds. It also delusions the so trendy Amazon cloud. Yes it might save some money/time to stand up a service, but what a performance penalty compared to some old school privately own hardware.

And a few others, adding a comparison with some high end server cpu

What's even more interesting regarding the frameworks comparison, is how they scale up if more cpu threads are made available, with the high performance Xeon processor, with 40 HT cores, vs 8 with the i7 cpu, candidate such as Spring, Grails or ruby on rails don't even offer double performance, while netty, using pure servlet or cpoll process requests 4x or 5x faster.

See full article here.