At two in the morning I was rapidly firing up EC2 instances in a desperate attempt to keep our platform alive. It wasn’t working. We were on track to losing our first big customer. We did. But six weeks later, we closed the deal.

This is a story about building a time-series database in lambda. About the agility being able to spin up 3000 machines in an instant can give you. About how you can use that to learn about the product you need to build. And about a team of three developers who no longer hide under their desks every time one of our customers treble in size. It’s also about that sinking feeling you get when you notice bits of your architecture randomly failing- and you realise there’s some important small print in the lambda definition of “scalable”. If you’ve ever wondered what it’s like to build and run your product on map-reduce in AWS lambda. This is the talk for you.