Nginx, erlang, golang, java http server benchmarks

timyang.net

34 points·hendler··6 comments

See relevant comments here http://www.reddit.com/r/programming/comments/a3lh8/c_erlang_...

I'd love to see node.js added to that.

A single instance of node.js running the sample hello world code from the site will not fare well as it will not be able to use multiple CPUs/Cores. Multiple instances need to be run with a load-balancer to make it comparable.

In my own previous benchmarks I've found node's performance to be close to Netty's when I ran 20 instances of node behind nginx on a quad core system.

I also found that there always was a single request which took several seconds. I suspect this has something to do with nginx/JMeter as I found similar behavior with nginx+php-fpm too.

As much as I appreciate the effort that goes behind benchmarks like these - it's important that these should not be taken as some sort of holy grail. Every project is different, and those factors are far more important than generic tests like this to decide on a particular technology stack.

Note: The article is a year old (It was posted exactly a year ago today).

The Go language benchmarks have probably improved considerably).

I'd like to hope that the Go performance improved, but I doubt it. My guess is that Go wastes most of its time in memory management and in goroutine scheduling, and I don't think either of those have improved significantly in the last year. I could be wrong, though.

Reply on news.ycombinator.com