Overprovisioning: A costly cloud computing mistake

The scourge of overprovisioning

And what are the DevOps teams buying too much of? Quite simply, they're too often purchasing too much cloud computing power. "People overprovision infrastructure, buying capacity just in case they need it, but the cloud allows you to spin up infrastructure as you go and as you need it," Adams says. It's an old habit that dies hard, as historically, IT departments have had to build systems that were big enough to handle peak demand. But that's not how you provision in the cloud. In the cloud, you provision based around average consumption, and then allow the cloud vendor to deal with the occasional spike. "Overprovisioning is a cost you don't need to incur," he says.

Getting aggressive with Elastic Compute Cloud reserved instances

What's the other expensive mistake too many AWS clients are making? "The second mistake is that people are a little too slow in buying reserved instances, which is Amazon's way of lowering the hourly cost of infrastructure in exchange for a prepayment, Adams says. Refusing to pull the trigger on purchasing a reserve instance is not only padding the pockets of Amazon, but it's hurting the bottom line of many AWS clients.

So, why aren't more people purchasing reserved instances? The problem is that figuring out what type of reserved instance to buy can be somewhat overwhelming, and clients are worried that they're going to make a mistake that might cost them more, instead of saving them money. "There are over 2000 different types of reservations you can buy, depending on the instance type, the availability zone you buy it in and the operating system, and each one of those reservations has a different break-even point," Adams says. And with so many different variables going into choosing the right reserved instance, many organizations are simply choosing not to make the decision for fear of making the wrong one.