Ruby Performance Roundup: Ruby 1.9.1 Real World Performance, GC vs EventMachine, Ruby Compiler
One argument for using Ruby 1.9.1 is the significanlty increased performance. New benchmark results come from running realworld, existing applications and pit 1.8.x, JRuby and 1.9.1 against each other:
We ported Acunote - our online enterprise project management and Scrum software to both JRuby and Ruby 1.9 and ran our set of performance benchmarks.
Ruby 1.9.1 and JRuby both provide significant improvements over 1.8.6, with 1.9.1 in the lead over JRuby, although there is some discussion in the comments about ways to improve JRuby's performance with some command line flags.
Performance improvements in 1.9.1 don't just come from a faster VM, but also from some of the new features. Muhammed Ali shows how to scale a web application using Ruby 1.9.1's Fibers. On the other hand, Muhammed also points out an issue with
Object#extend leaking memory in 1.9.1.
Meanwhile, 1.8.6 is still the only option for some projects, due to some missing libraries on 1.9.1. Because of that, there's a lot of interest in fixing some bottlenecks in 1.8.6. Joe Damato has been investigating some of the issues over at his blog. For instance, he investigates the story behind --enable-pthread and why disabling the setting brings 30% performance gains. In another article, Joe and Aman Gupta investigate a problem with the Ruby GC - and came up with a tiny patch that fixes some problems with the GC and EventMachine:
* Speeds up GC by 2-3x because of the huge decrease in stack frame size.
* Fixes an open bug in EventMachine where using threads with Epoll causes lots of slowness. The reason is that each thread will inherit an ~800,000 byte stack that gets copied in and out every context switch.
* This results in an increase from 500 requests/sec to 7000 requests/sec when using Sinatra+Thin+Epoll+Threads. That is pretty ill.
Finally, Viktor Hokstad has been writing a series on compiling Ruby for some time now. A recent entry talks about some of the problems making Ruby fast and possible optimizations.
Tom Gilb & Kai Gilb Jan 26, 2015