Billion Events Per Second, Millisecond Latency: Giga Streaming Analytics
On a scientific paper on Hazelcast Jet, its architecture based on symmetric, data-local, non-blocking distributed event processing.

Marko Topolnik, PhD, has been a Java professional since 2001. His current position is in the core team of Hazelcast Jet, where he co-wrote the core execution engine based on coroutine-like suspendable code that runs many concurrent tasks on a fixed thread pool. Marko is also an active contributor on Stack Overflow on the kotlin-coroutines tag.

On a scientific paper on Hazelcast Jet, its architecture based on symmetric, data-local, non-blocking distributed event processing.

Same problem from the opposite angle: what can we do to help Hazelcast Jet achieve the best performance available on a JVM?

This is a followup on the series analyzing performance of modern JVMs on workloads relevant to the use case of real-time stream processing.

Let's consider the experimental low-latency collectors in this round, since their very short GC pauses have no effect on the test result.

Hazelcast put new Java technologies to the test on workloads typical for their distributed stream processing engine, Hazelcast Jet.