I was initially playing with Erlang, talk about good times, and I just got
frustrated trying to do things I can do easily in Ruby and decided to fall
back to Camping which I like a lot.

Then I came across Celluloid (http://celluloid.io/) and with it on JRuby
you can pretty much take full advantage of all your CPU cores.

Plus it does the 'let it fail' actor model, so when your Camping app
crashes, it just restarts the actor again.

This can make for a really robust webapp with all the benefits of running
within a JVM.

I then saw Reel-Rack which made running Camping easier, but I had the
question of, how good would the performance be with all the overhead of JVM
- JRuby - Celluloid::IO - Reel - Reel-Rack - Camping.

That's when I started putting together the benchmark project just to see
how that setup would look against a MRI version of my webapps.

The simple tests don't really do it much justice, I'm starting to play with
DCell and ZeroMQ, so I might try to come up with some examples where JRuby
can use all the cores and see how MRI does with the same code.

I also tried to play with Rubinis but ran into install issues with some
Gems so I decided to stick with JRuby.

I'm also on SmartOS in a Zone and KVM for Linux, so the tests might be
completely different on a pure linux install on the same hardware, for
better or worse.

Would be interesting to see but I'm not going to rebuild my server to find
out, maybe I'll try it in AWS or something.

On Mon, Nov 18, 2013 at 2:53 PM, Tim Uckun <timuc...@gmail.com> wrote:

> Those are fascinating and very surprising results.
> I thought for sure trinidad would rule over all of them.  The performance
> of Puma is also very surprising.
> So do you think there is any reason to use jruby at all given your
> benchmarks. MRI seems to be pretty good.
> _______________________________________________
> Camping-list mailing list
> Camping-list@rubyforge.org
> http://rubyforge.org/mailman/listinfo/camping-list
Camping-list mailing list

Reply via email to