evolution of hosting - cloud computing
Posted in hosting on September 29th, 2008OK you’re a startup without the money or patience or expertise to setup your own servers in a way that will enable you to scale to a spike in demand. You hope you’ll experience growing pains, but worry about it too.
You pay as you go in the cloud computing model. Define server “instances” and create and kill them as needed. Defining server instances mean to decide what applications go on the server, and the configuration options you’ll need. I’m oversimplifying it a bit, but in general you can automate the process of spawning and reaping server instances based on traffic levels. There are some nice tools available to help you set triggers, specific amounts of traffic that will cause a new instance to start up, or go away.
I was oversimplifying it, that’s true. For webservers this model is just about as easy as it gets, because many others have defined server profiles you can choose from and use. But obviously a database server is going to pose additional challenges, right? You can’t just pull up more database servers without some plan for replication. You must plan in advance to scale… I mean start with a single server, be prepared to setup a backup DB server as a slave, proceed to having multiple database servers, some replcated servers dedicated to read operations and some sensible strategy for passing update data around. But hey, if you actually make it to the end game in your plan you must surely be hitting a home run with your website!
Amazon is the mindshare leader here with their Elastic Compute Cloud (EC2) service, but there are plenty of rivals cropping up everywhere. Amazon gives you the nice benefit of being able to use their S3 storage service in conjunction with EC2 without incurring a data transfer fee. Same with their queueing service SQS.
Why would you consider this? You’re a startup, as I said, and your budget and time are both limited. Pay by the hour as you need - what could be better?


