How many servers do you need for your cloud?

cloud servers DXC Blogs

One of the neat things about cloud, as a user, is that you don’t have to worry about how many servers you need. Need some more? Just spin them up. No fuss, no muss.

But what about when you’re the cloud architect? You know — the guy or gal who must buy the servers that make up the cloud? Well, in the case of Amazon Web Servics (AWS), they turn to artificial intelligence (AI).

AWS CEO Andy Jassy told attendees at the Pacific Science Center’s 14th Annual Foundations of Science Breakfast that Amazon has been using machine learning to anticipate demand for its servers. “One of the least understood aspects of AWS is that it’s a giant logistics challenge, it’s a really hard business to operate,” he said.

This is true of any cloud operation. As a company turning to public cloud, you are indeed replacing capital expenses (CAPEX) with operating expenses (OPEX), but that’s not how it works for a public cloud company. Or, for that matter, your company if you’re building your own private cloud.

The advance of virtualization technology has helped hide the CAPEX expense. Virtual machines (VM)s enabled users to get more production from bare metal servers. Then more recently, the rise of container technologies, such as Docker, allowed businesses to get even more workloads running on the same servers.

But, at the end of the day, no matter how powerful your virtualization magic is, you ned to buy more servers. How many? Well, you can’t solve that question by simply buying truckloads of servers to make sure you have enough. Not to mention, your CFO might have a thing or two to say about that plan.

Even AWS, the biggest of the public clouds, can’t do that. “Every single day we add enough new servers to have handled all of Amazon as a $7 billion global business,” said Jassy. That’s a lot of servers.

As Jassy explained, AWS uses a machine-learning based forecasting model to make CAPEX-buying decisions. In the model, AWS considers such factors as the long enterprise sales cycle and that AWS customers tend to demand more resources once they’ve gotten used to placing workloads on AWS.

Google and Microsoft also use similar methods in their businesses.

Unfortunately for us, none of these companies is sharing the methodology. Open source may be well and good for the technical side — the cloud couldn’t exist without it. But, when it comes to business processes, mum’s still the word.

That’s a pity if you want a hybrid or private cloud, since this information would make it much easier for businesses to decide how to build out their cloud server farms. Over time, I hope AWS, or another of the other major public cloud powers share, if not their data, then at least their methodology.

That would make life so much easier for us.


Why advanced analytics needs the cloud

Reducing total cost of ownership with cloud ERP solutions


  1. Thank you for your support to us for this topic. I need this information and I feel very happy to getting here only. I am very satisfied with your site and also I like your site it’s very interesting one.

Speak Your Mind


This site uses Akismet to reduce spam. Learn how your comment data is processed.