Significant performance degrade when updating aws lambda runtime from java 11 to java 17

Question

Hi

I'm trying to update a java11 based lambda function to run java17, but I experience a doubled execution time for simple function that just reads a large object from S3.

I posted a lot more details on [stackoverflow](https://stackoverflow.com/q/76761254/1714215) but have not yet gotten any responses, hence this crosspost.

Is anyone else experiencing similar performance degrades on java-17?

Best regards Jens

Accepted Answer

The JVM configuration for tiered compilation was changed in the Java 17 managed runtime.

You could change it back to the previous configuration with a environment variable.

`JAVA_TOOL_OPTIONS` set to `"-XX:-TieredCompilation"`

You can read about the changes on the [Java 17 AWS blog](https://aws.amazon.com/blogs/compute/java-17-runtime-now-available-on-aws-lambda/)

Answer

Hi, I would personally start by tracing garbage collection and compare between v17 and v11. There has been lots of changes in this area between the 2 versions. In my experience, some GC changes good for most use cases are (very) bad for some. You may be in this situation.

I am saying this because I have been in projects when some code had to be rewritten to fight adversarial  GC activity (blocking)  due to too much memory allocation / release in the Java classes for class and method variables. We had to minimize this activity to get back to correct performance. Form your Stackoverflow post, you seem to read / write several GBs of data. So lots of chances for similar big memory activity and GC.

So, to trace GC: https://aws.amazon.com/blogs/architecture/field-notes-monitoring-the-java-virtual-machine-garbage-collection-on-aws-lambda/

About the evolutions: https://blogs.oracle.com/javamagazine/post/java-garbage-collectors-evolution

Best,

Didier

Significant performance degrade when updating aws lambda runtime from java 11 to java 17

Relevant content