Faster Kafka Without Changing Any Code

One of the primary reasons Kafka is built on Java is its incredible scalability, ease of use, and "write once, run anywhere" capabilities.

However, the JVM, a managed runtime environment, both helps and hinders in achieving the best possible Kafka performance.  Aspects like garbage collection pauses can make it hard to meet defined SLAs.  JIT compilation can deliver better overall performance, but your Kafka nodes take time to warm up.

In this session, we'll explore at a technical level how Azul has been addressing these limitations and building a high-performance JVM that delivers fantastic results for Kafka.

Specifically, we'll look at:

• How can GC operate completely concurrently with application threads whilst still moving live data within the heap to eliminate fragmentation?
• What can be done to improve the level of optimisation of code generated by the JIT compiler
• Solving the warmup problem.  There are several approaches to this, and we'll provide a quick comparison of the advantages of each and where they can be used effectively
• Finally, what if we decouple the JIT compiler and move it to a JIT-as-a-service (or a Cloud Native Compiler)? How can this further improve Kafka applications?

The results can be impressive, with benchmarks showing over 50% more transactions per second in certain situations.

This session will cover low-level technical details to help you understand your options when deciding how to optimise the Java environment for your Kafka deployments to maximise performance.

By attending this session, your contact information may be shared with the sponsor for relevant follow up for this event only.