Consider that 37Signals probably has a much larger dataset and is still happily scaling vertically due to Moore's law. http://highscalability.com/blog/2012/1/3

Consider that 37Signals probably has a much larger dataset and is still happily scaling vertically due to Moore's law. http://highscalability.com/blog/2012/1/30/37signals-still-happily-scaling-on-moore-ram-and-ssds.html http://highscalability.com/blog/2012/1/30/37signals-still-ha... It's probably the cheapest, quickest, and least risky option if your engineers' time is expensive (which it most likely is). Of course, you'll ultimately also want to find bottlenecks and tune the DB and cache the hell out of everything, but scaling vertically will at least buy you some good time so you can do that in a more relaxed time-table. Related links: http://37signals.com/svn/posts/3202-behind-the-scenes-the-hardware-that-powers-basecamp-campfire-and-highrise http://37signals.com/svn/posts/3202-behind-the-scenes-the-ha... http://37signals.com/svn/posts/3090-basecamp-nexts-caching-hardware http://37signals.com/svn/posts/3090-basecamp-nexts-caching-h... http://37signals.com/svn/posts/3112-how-basecamp-next-got-to-be-so-damn-fast-without-using-much-client-side-ui http://37signals.com/svn/posts/3112-how-basecamp-next-got-to...