Select Page

Last time, we mentioned our mission to dramatically cut the cost and complexity in global deployment of containers to delight your users.

How big a problem is cloud spending? Let’s take a look.

Current spending

How much are we all spending on the cloud? Gartner estimates that public cloud revenue in 2020 will break a quarter of a trillion dollars. And it’s early days with immense growth still ahead! It’s not all roses though. Enterprises know they’re wasting at least one third of that spend, and in the near future the industry will be wasting $100 billion each year!

This obviously isn’t sustainable. It’s no wonder that three quarters of cloud users see optimizing cloud costs as their most important initiative, as reported in Flexera’s “2020 State of the Cloud” report, even as migrating more workloads to the cloud is next on their list.

What’s causing this spiraling waste? There are many factors from small scale to large. From simple VM sprawl, to cluster-induced waste, to wide-area structural waste that cloud products foist on you due to making gold standard architectures such as “multi-region active-active” deployments an application developer-level concern.

Resource sprawl

At the simplest level, clouds are prone to resource sprawl. People spin up resources, like virtual machines and storage, walk away, and get billed indefinitely for it. The cloud has made it trivial to allocate resources and harder to tell when they should be turned off.

To start with, even choosing resources is tricky. Say you want a VM. What instance type and size do you need among a choice of hundreds? Which processors and how much memory? Should you use 1 bigger machine or 2 smaller ones? In which zone? More than one zone? More than one region? Where’s your data and which network charges will you continuously trigger based on computing in that location? Should that be a reserved machine with a lower price based on correctly guessing our future needs and committing to it?

It’s hard to make the right choice and painful to undo as time goes by.

Over-provisioning is common because who wants to have to come back and increase the size of the machine later if you’re wrong about your needs. Having made a choice, we’re all loathe to go through the process again especially given whatever else we’ve layered on top. After a while, it’s perhaps not even worth the risk as it’s often not obvious who will start yelling when the instance is removed!

For this and many other reasons, as enterprises mature in their cloud usage, they increasingly shift to Platform as a Service (PaaS) products such as containers and databases as a service. Next time we’ll touch on the rise of containers, and cloud structural sprawl.