Tech Lover 9

In the early 2000s, an online store was gaining huge popularity, so much so that they found it increasingly hard to manage Christmas shopping with their current server base, needing to increase it’s size and capacity every year. This led to a much bigger problem – now that they had so many servers to cater for Christmas shopping peak-hour demand, what to do with all this capacity for the rest of the year? So in November 2004 Amazon launched its first public service, the Simple Queue Service and in 2006 started its now highly prominent Amazon Web Services, including EC2 and S3.

While many Amazon insiders debunk this story as a myth, claiming that the AWS was always supposed to be its own business, it illustrates several key factors that all IT professionals need to think about before debating on whether to move to the cloud.

Grid vs Cloud

The first key here is to note the difference between a grid and a cloud. If Amazon had worked on a grid in the fashion of IBM before it, it is estimated it would have run out of ‘excess capacity’ in 15 days of launch, instead of the estimated 2 months for a cloud. This is because with a grid, you can scale, but only in discrete units. So, giving out a server to everyone who asked for one, even simply for storing one file, would have meant the end of AWS before it even started – either because they would run out of computing power, or because they would be so overpriced, that they would be beyond the reach of any regular computer user.

Crucially, the stress is on minimal wastage. If someone wants an extra MB, you give them an extra MB, and if someone wants an extra PB, you can give them that too, because you’re not limited by the size of the server. Everything becomes cheaper – Amazon could use the servers it had extra entirely, using up every drop of computing juice they could muster.

Having talked about the differences, it is important to note that all the advantages of the grid were incorporated into the cloud. Clouds are just as scalable and reliable. They just simply come without the need for batching, or pre-defined, specific beginning and end points.

Another key lesson here is on-demand availability. The problem Amazon faced was seasonal. They had to find a way of using their complete server power no matter what time of the year it was. This would have led to a direct conflict of interest between the AWS and Amazon Retail, had it not been for a cloud. Now, they could transfer power wherever it was needed, without the assumed additional overheads of context switches and intellectual power, since everything now is just one single unit. You can charge people for each and every hertz of processing they use up, and shift the whole thing to someone else when they’re done.

As an IT administrator or a business owner, here’s what you can learn from the makers. It would almost always make sense to go with the cloud, unless you are big enough to go out and buy and operate your own computer city, because in the long run, they cause lesser headaches, and allow you to move beyond what you thought possible.