Knowledge Center

Limitations of PaaS


Platform as a service is really for you guys actually. Most of you are developer types. You’re not infrastructure guys. And so, Infrastructure as a Service still means you have to know all this stuff. What if you don’t know how to deploy a Postgres server and configure it to run in a fast way. Postgres, which by the way, we are the Southeast Asian press contact for the Postgres project. We’re big Postgres fans. Postgres out of the box is configured to use absolutely the minimum amount of resources possible. You’ll have no problem uploading and installing a stock version of Postgres on a Raspberry Pi and it’ll just run fine. That’s not good if you want to be performant. To use Postgres well, you got to know where to plug in some options and use the memory and use the CPUs and stuff like this. If you’re not a Postgres guy, you’re not an Ops guy, you not going to know how to do that. So services like Heroku. They’re the first big one and they were initially focused on the Ruby world and now they support different software stacks inherently, like pHp. So instead of utility computing, they do utility operations and this is actually from their website. I think it’s a pretty good description. It’s showing here’s all the things that you got to do for doing infrastructure, here’s all the non-functional stuff. That’s the developers over here, and then it drags the slider over to the right and you’re on Heroku, this is all you have to think about, and all the rest of the stuff is handled for you.

Downside of Platform as a Service is, you’ve just got vendor lock-in back, because when you’re deploying into Heroku or deploying into or any of these Platforms as a Service, you’re now having to design and code against a proprietary API. So Openstack is trying to save you from that, but it doesn’t quite work in that way The neat thing about infrastructure as a service is you have this utility computing. It’s literally like plugging into the wall socket and getting computing power. You get disk space you get memory you get CPU computations. And you get network. And you just get it by this incremental amount as much as you want or as little as you want whenever you want. It’s fungible. You can use those compute resources to do whatever you want and the way you want it. You can optimise however you want. When you’re Platform as a Service, you’re on Heroku, and let’s say that you’re using their Postgres configuration, you’re deploying in that. If your application happens to be very database intense and the way you’ve differentiated against your competition is by having very fast database access, you’re kind of screwed. Heroku will scale to a certain extent but Heroku’s database configurations got to be defined for the most common denominator, greatest common factor, of its user base. They’re not going to spend any time optimising for the way your system uses the database

Back to Knowledge Center