Programming AWS Lambda

Foreword

AWS Lambda—and both serverless and backend as a service in general—have had a hugely disruptive effect on the software industry. They’ve greatly improved the productivity of millions of developers by eliminating many of the hassles, costs, and “undifferentiated heavy lifting” of dealing with servers, from security patching to autoscaler tuning. More importantly, though, serverless has changed the very definition of an application, from a blob of code that we drop onto server farms into a configuration of multi-tenanted cloud services that we orchestrate with code in the sky. Serverless is the next stage of cloud evolution—just as it once felt impossible to build a company without its own data center, it’s quickly becoming possible for companies never to own a server. It’s a fascinating transformation to watch and be part of!

When I was first coming up with the ideas that eventually became AWS Lambda, I had many discussions with the AWS leadership team about the risks and opportunities. The opportunities were massive—the chance to reimagine how compute and applications were constructed and to change the very nature of software development in the cloud. But the risks were equally great. The innovative “spark gap” of Lambda was high, and crossing that chasm required a lot of energy: to gain the advantages of serverless—pay per use, per-request scaling, built-in fault tolerance, and so much more—we’d have to ask developers to give up conventional server- and container-based ...