Platform as a Service: A Market Analysis

The thing that current PaaS solves

There are quite a few companies trying to compete for a developer friendly hosting/deploy platform. Among those are heroku, openshift and cloudfoundry, to name the biggest players.

PaaS essentially solves the struggle to deploy apps (typically web apps) to the internet. What you could have done with a few build scripts, ansible or chef is now possible without any hassle, especially because it solves:

• No downtime, also during deploy (depends on hoster).
• Automatic app detection, typically with heroku buildpacks, even dokku uses those. Which means no configuration at all.
• An integrated backup solution - backup is not an issue anymore.
• Multiple staging/deployment "clusters", that one can target (see cloudfoundry)

I know, some of the PaaS solutions may not solve all of those points, but they are basically things that are not so easy to solve with "just another deploy tool".

It is important to note at this point, that PaaS are not an option for a lot of in-house apps, because even if they are created open source - like openshift and cloudfoundry - it's not that easy to set them up, because your own cloud still requires a lot of ressources and maintenance.

Why is PaaS not enough?

The big problem with PaaS is basically that they don't really give you power over the servers you are using. The most visible manifestation of that is that almost nobody uses static files with heroku (and there's probably the same issue with openshift).

I really don't want app development - I want software development, where you as a developer have control over the machine. This means configuring nginx and so on. Being limited as a web developer reminds me too much of the old PHP shared hosting times. What if you realize that you can gain a lot of speed by using a certain cache, but your PaaS doesn't support that? Maybe it doesn't support memcached?

But really - PaaS could also allow more configuration

I think one of the problems that PaaS has right now is that people thought too much about billing customers. I know cashflow is important - but having happy customers also allows for that. They invest a lot into new services, but not into a multitude of configurations of services. I think a combination of a depl style configurations and a good docker service with backup capability and HA load balancers could be a really interesting new PaaS for example.

Do you agree? I would really like to have your opinion on depl and its approach. Especially the configuration API.