Last night, Google launched their Amazon EC2 competitor, App Engine.

For those of you that don’t know - Amazon EC2 is a system whereby you can upload a machine image that you can then create literally hundreds of virtual images from for very little cost.  Accompanying this is the Amazon S3 service which is essentially a huge virtual storage device in the cloud - priced in a very reasonable pay-as-you-go style.

The cool thing with EC2 is that it is all controllable with an API - so, for instance, you could potentially have a web application that monitors it’s own load, and scales out new servers as required fully automatically - you only pay for what you use.  This is great for new start-ups who aren’t really sure if their new super-duper social site is actually going to get no visitors, or become the next Facebook.

Google App Engine is a little different.  With this, you are given a pre-prepared machine image which is basically a web server rigged up to use Python (with other platforms potentially coming soon) and a Google driven data-store, which is bound to be a bit speedy.  Initially this might seem a little annoying as a CFML developer until you check out the free account.

As a non-paying App Engine subscriber you are somewhat limited on what you can do - but as Google describe it, you have enough for half a gig of storage and around 5 million page views a month.  Now, thats a pretty busy website in my eyes.  So, if you have a busy static website, a load of files requiring storage, or are handy with Python it’s worth a look - you could end up with a free web server.

On the other hand though, as a CFML developer, you’ll need to look at EC2.  With the advent of the open-sourcing of Bluedragon, we are now in a position where we can build a no-cost server image (comprising of Linux, MySQL, Apache, Tomcat and Bluedragon) which we can then deploy to EC2 and scale for very low cost and end up with a supremely scalable, and performant cluster of servers.  Best of all, if you don’t use it much - it doesn’t cost you as much, unlike having a rack of under used servers somewhere in a data-center.

So, come the revolution, we can all be building our infrastructure from nothing but the cloud, and scaling as far as we want, and best of all, for no cost but still using CFML.  Obviously Amazon EC2 is the only choice at the moment for cloud CFML computing, but I’m sure Google will make their system a little more customisable over time.