One of the main benefits and selling points of serverless solutions like Knative is that it saves CPU cycles and RAM consumption (and by extension, money) by only using resources when they are actually needed. Because of that, we would semi-automatically infer that using a serverless approach when architecting our applications is the way to go if we want to be efficient and do our planet a favor by saving energy.

But wouldn’t it be nice to, actually, have some data to support this statement? Until very recently, we could only guess. But technology is advancing fast and now we have tools to observe and test this hypothesis.

In this talk we will show if, and if so how much, serverless can save not only in computing power but real energy as well. We will do this by actually measuring energy consumption of nodes and workloads with Kepler and shed some light on this topic to figure out if our assumptions are true, or just a myth which needs to be busted.