AWS Uses Machine Learning To Decide How Many Servers To Buy

AI being put to good use here! Technology giants have been increasing their reliance on machine learning to make certain decisions, and AWS is using this technology to make capacity planning decisions.

Basically, how many servers to buy, and when to buy them.

At the end of the day, Amazon Web Services is not only doing machine learning research for its customers, but it has also developed machine learning driven models on how to forecast demand for its services, and this helps the cloud giant deploy infrastructure accordingly.

AWS CEO Andy Jassy talked about this practice in front of an audience at the Foundations of Science Breakfast by the Pacific Science Center, detailing how the company has to buy an enormous amount of servers regularly:

“One of the least understood aspects of AWS is that it’s a giant logistics challenge, it’s a really hard business to operate. Every single day we add enough new servers to have handled all of Amazon as a $7 billion global business.”

And although the report above does not go into details about what kind of input data the company feeds into its machine learning algorithm in order to forecast demand, but it looks like that one of the primary data sources is from its cloud sales team.

Apparently, it can pick up signals from the processes that its sales teams follow to forecast demand.

Which is rather impressive, as enterprise sales cycles are known to be notoriously long. And not just that, most new customers start slow on AWS and then accelerate their usage as they see more benefits, and this can lead to spikes in demand if they move faster than anticipated.

Machine learning also helps the company to determine where to store excess components for its datacenters, in order to react quickly when more capacity is needed in a given region.