At what point does scalability become an issue for a Rails application deployed on Heroku?

Question

Ruby on Rails has largely been faulted for it's lack of scalability options, and the alternative is generally to migrate to some form of a Java webapp or something similar. But all of the concerns seem to be very arbitrary at best, in a way.

Are there any concrete numbers for when a certain application needs to be rewritten in a different language? Heroku provides various scaling options with the numbers of dynos available to the app, but at what point will diminishing returns, if any, be evident? Or at what point will the cost of having so many dynos outweigh the costs of simply writing a new app?

How many active concurrent users can I expect to be able to support without suffering performance issues on the basic free hosting plan at Heroku?

Answer 1

The "Rails doesn't scale" myth simply won't die ...

Quoting Twitter's lead architect:

The performance boosts associated with a “faster” language would give us a 10-20% improvement, but thanks to architectural changes that Ruby and Rails happily accommodated, Twitter is 10000% faster.

I can't give you a fixed number, but it is almost absolutly guaranteed that you run into database or architectural bottlenecks long long before Ruby or Rails are starting to become bottlenecks.

Teflon Ted has created a nice collection of blog posts regarding this topic.