Introducing Phusion Passenger 5 beta 1, codename “Raptor”
blog.phusion.nl
blog.phusion.nl
We still use version 4 and benchmark it with unicorn and the likes, and we found no noticeable different on performance, except the fact that it's easier to configure and manage.
I had been following Raptor and actually had no idea it was by the Phusion folks. Guess I didn't pay close enough attention! Very excited to test the betas and see how it performs across our network.
In benchmarks with customer's systems against Passenger4, Puma, and Unicorn, we have not seen anything that would suggest a general-case migration to any of those would be advisable for apps without a specific workload reason.
We have a number of folks on the Enterprise product, primarily for zero downtime restarts. I think "Immutable Infrastructure" plus load balancer support would negate the need for this, if you've got the orchestration in place. I'm curious how Phusion responds to that becoming more and more popular/prevalent.
We had 100 req/sec on our 2gb test server.
We got 170 req/sec without any tuning by using the same config
Im sure i can tune it little more to get it to 200 req/sec, which is a lot improvement to previous version. Well done guys!
I'm disappointed they're dropping the Raptor name, it just sounds so much better than "Phusion Passenger".
[edit] Actually trying it out on my dev machine right now. I am getting a random error that I normally don't get with thin server. Doesn't play well with pry either right now. I'll wait a bit until it goes out of beta.
I also don't think the comparisons Raptor is doing are fair. You're comparing an application server with a cache against just application servers. Also, I would expect Raptor's memory usage to be higher once that cache starts to fill up, or maybe you've created a super small cache just for "Hello World" apps.
People aren't avoiding Passenger because "it's too easy to use", a lot of people avoid Passenger because you're a weird company. Application servers are very important, perhaps too important to be tied to a commercial entity.
I applaud you on the work (code-wise) you've put into this and I hope the entire Ruby community as a whole benefits from it, like you claim we have with older versions of Passenger (which kinda seems like revisionist history to me, but oh well).
You don't remember "I love Unicorn because it's UNIX?" http://tomayko.com/writings/unicorn-is-unix as well as GitHub's postings about using Unicorn?
I admit this particular campaign makes me a bit... unhappy. But every open source project does marketing, explicit or implicit.
I think exactly the opposite. I think apps servers are too important to be left without financial support.
Remember when Zed Shaw abandoned his great work on mongrel? I don't think he'd do the same if he had thousands of paying users, and maybe we'd still use (maintained and improved) mongrel today.
I absolutely cannot justify spending $30/server month on passenger - that's more than the boxes I'm running it on cost - and zero downtime restarts are a must-have for me.
Depends though, sometimes this is a nifty alternative, in other cases it may not be a substitute at all.
Kudos to them!
It's much faster. Learn more at http://www.rubyraptor.org/ and http://www.rubyraptor.org/how-we-made-raptor-up-to-4x-faster...
It has better tooling for helping you to identify and solve problems.
Example Procfile configuration: https://github.com/phusion/passenger-ruby-heroku-demo. Note that the demo locks down to Phusion Passenger 4 by default. If you want to try 5, don't forget to bump the version number in the Gemfile.