While the array of cloud platforms (Platform as service) has increased significantly recently, its still a hard job to find the perfect host for your website. The obvious advantages of cloud platforms over dedicated servers start with cost and extend to ease of maintenance. For instance, on EC2 it takes about 30 min or less to fire a new server instance and the Amazon’s ability to charge its customer on per hour basis, makes EC2 a very attractive option. AutoScale is another great add-on to EC2 platform and lets the consumer automate the computing power usage based on user load.
We recently did an extended research on available cloud platforms for a client. The requirements were however no straight forward. Client’s top priorities were ability of platform to scale automatically, ease of maintenance, and performance of infrastructure. Then, to make the task challenging and exciting, the client wanted a geographical fail-over ability in case of a disaster. Amazon alone had 2 major outages in year 2011 so users are quite concern about business continuity in time of a disaster.
We evaluated many options starting from Amzon EC2, RDS, Heroku, Bluebox, RackSpace, High Velocity, Xeround etc and concluded that only one provider will not be able to answer all questions from our client. So we divided the architecture into App layer and DB layer and choose different hosts for each layer. In such a division the biggest concern is network latency but interesting fact is most cloud platforms are built over EC2 with additional wrapping so within EC2 network latency is not huge at all. We ran some test with Heroku (app) and Xeround (DB) and found network latency to be less than half a sec per transaction.
You can find below a presentation we put together after our analysis of different cloud hosts. There is a slide that links to a Google spreadsheet with calculation of estimated cost for our proposed infrastructure. Feel free to drop a comment to correct me or ask me