Hazelcast Jet: Low-latency Stream Processing at the 99.99th Percentile

TL;DR

Hazelcast Jet is an open-source distributed stream processor optimized for millisecond latency at the 99.99th percentile, capable of handling out-of-order streams with high throughput and strong processing guarantees.

Contribution

The paper introduces Hazelcast Jet's design for achieving ultra-low latency and high throughput in distributed stream processing, including handling out-of-order data and ensuring exactly-once processing.

Findings

End-to-end latency in milliseconds

Throughput of millions of events per CPU-core

Effective handling of out-of-order streams

Abstract

Jet is an open-source, high-performance, distributed stream processor built at Hazelcast during the last five years. Jet was engineered with millisecond latency on the 99.99th percentile as its primary design goal. Originally Jet's purpose was to be an execution engine that performs complex business logic on top of streams generated by Hazelcast's In-memory Data Grid (IMDG): a set of high-performance, in-memory, partitioned and replicated data structures. With time, Jet evolved into a full-fledged, scale-out stream processor that can handle out-of-order streams and exactly-once processing guarantees. Jet's end-to-end latency lies in the order of milliseconds, and its throughput in the order of millions of events per CPU-core. This paper presents main design decisions we made in order to maximize the performance per CPU-core, alongside lessons learned, and an empirical performance evaluation.