Parse, the backend-as-a-service for mobile and web apps, is shutting down, and Marco Arment explains one of the reasons he doesn’t use those sorts of things:

The short answer is that I can’t afford to — for my business models to work, I need to keep costs very low, a discipline I’ve built over time as one of my most important professional skills. […]

For whatever it’s worth, running your own Linux servers today with boring old databases and stable languages is neither difficult nor expensive. This isn’t to say “I told you so” — rather, if you haven’t tried before, “You can do this.”

I agree that getting a Linux box with one of the cloud-hosting providers is very easy, with numerous vendors and 3rd party tutorials provided to get you up and running. However services like Parse can offer peace of mind if you suddenly get 1,000,000,000 users and need to scale. Instead of having to learn how to deploy very complicated load balancers and shards, you can pay Parse to work it out and keep the lights on. Of course, that peace of mind can evaporate suddenly, as it did here, if they cannot keep the lights on with what you pay.

I predict the problem with Parse was also the reason it was so successful: the barrier to entry was so low, and the ratio of paying customers to free-tier customers was probably too low. What I don’t understand is that Facebook could probably afford to run Parse at a loss forever, and just keep tabs on up-and-coming apps.