If you have ever run Java inside a lambda function on AWS, you will have noticed the quite significant cold start times that comes with spinning up the JVM environment. In this post, I will discuss some different tricks you can use to minimize these cold start times.

The problem with cold starts arises when there are no “warm” lambda available to handle an incoming request, which usually happens whenever an endpoint experiences a large and sudden spike in traffic. The most commonly occurring scenario when this happens is probably when an endpoint goes from no traffic at all in a while (and thus having no warm lambdas ready) to suddenly having one or more incoming requests to serve.