Are there any performance/functionality differences for AWS Lambda functions written in node.js/Java

Question

I am planning to use AWS Lambda for the backend of an app. I am more comfortable with Java compared to node.js but I see Lambda functions in node.js are more popular than Java. Are there any performance differences between Java based and nodejs based lambda functions?

Answer 1

While Java has a slower startup time, it's runtime performance is a lot better than the one of node.js or Python. That means Java Lambda functions make sense if:

1. They are Invoked regularly - AWS Lambda will reuse it's initialized containers and you can benefit from the better performance without the cold start latency

and/or

2. They have a long runtime - If your function is running e.g. for 1 or 2 minutes, you do not really care about 5 seconds of startup if the function itself is running an order of magnitude faster than node.js

You may want to take a look at this github project that compares the cold start times.

One big factor of Java's bad cold starts is probably the fact that Java projects are often a lot bigger than e.g. a node.js function. In any way you should always try to keep functions as small as possible to reduce the initial latency.

And of course performance shouldn't be the only factor when choosing your programming language. I personally think node.js is extremely convenient when working e.g. with JSON data, this is why I use it in most of my functions.

Answer 2

Java functions on Lambda generally require more resources and take much longer to respond from a cold start, when compared to NodeJS functions. Otherwise there is no difference.