Hazelcast, the leading in-memory computing platform company, today announced the general availability of Hazelcast Jet, the only streaming engine with no external system dependencies. The result is the industry's fastest stream processing engine that dramatically simplifies implementation from the smallest to largest deployments.

Whether deployed in constrained environments, such as IoT sensors, or cloud-scale applications, Hazelcast Jet ingests, categorizes and processes vast amounts of data with ultra-low latency to support continuous intelligence practices.

"SigmaStream specializes in high-frequency data and works with some of the world's largest companies that operate in the most constrained environments. By integrating Hazelcast Jet's high-performance streaming engine with our Hummingbird visualization and processing platform, we process high-frequency data from dozens of channels and address inefficiencies in real-time," said Hari Koduru, CEO of SigmaStream. "The performance and optimization at such a fine level enable SigmaStream's customers to shrink the time spent on a project, ultimately saving them millions of dollars."

Single System Design

Normally, deploying other streaming engines requires enterprises to invest the time and endure the complexity of integrating multiple components from different sources. For example, a Flink implementation would necessitate integrating a combination of Kafka, ZooKeeper, RocksDB, Hadoop File System and resource managers to ingest, categorize and process data.

Hazelcast Jet significantly simplifies deployment because it is a single, lightweight system that elegantly addresses a complex set of architectural requirements. Hazelcast Jet's unique single-system design enables rapid time-to-value, eliminates costs and complexity associated with multi-component architectures, and reduces the need for multiple skill sets.

Industry's Fastest Streaming

Internal benchmarks demonstrate Hazelcast Jet's ability to maintain millisecond speeds at extreme scale, where other open source-based projects drop into the seconds. Hazelcast Jet maintains its ultra-low latency, regardless of scale, due to a distributed architecture and in-memory processing.

"Hazelcast has once again delivered a powerful leap forward for the industry, this time by radically simplifying how stream event processing is implemented," said Kelly Herrell, CEO of Hazelcast. "Time is money, and the ability to process data at the moment it is generated - wherever it is generated - produces measurable business benefits whether at a financial trading desk or edge-based sensors. When time matters, companies choose Hazelcast and now they have a compelling and flexible streaming solution for fast data processing in Hazelcast Jet."

Importantly, Hazelcast Jet delivers low-latency performance regardless of scale, whether running at the IoT edge in small-format hardware or as a massive cluster in data centers and clouds.

Run Anywhere

Hazelcast Jet's architecture is simultaneously lightweight and highly scalable, allowing it to run wherever customers need high-performance stream processing. Its small file size and architecture provide numerous deployment options, including in Kubernetes microservices environments, private data centers, public clouds or embedded in applications.

Furthering the deployment simplicity of Hazelcast Jet, it is Kubernetes-ready to support containerized workloads and validated to run in Pivotal Cloud Foundry and Red Hat OpenShift cloud environments.

Elastic and Resilient

As workloads increase, Hazelcast Jet's clustering model scales up and down without job interruption.

Hazelcast Jet clusters can be taken offline without losing data and jobs can be upgraded without interrupting processing, a significant benefit for long-running continuous streaming applications. In the event of an outage, in-memory data replication provides a robust yet performant means of fault tolerance, with Hot Restart for fast recovery. The in-memory data can also be continuously persisted to disk for maintenance shutdowns or lights-out events.

Machine Learning Modeling

Whereas most streaming engines use batch processing to manage data, Hazelcast Jet is capable of processing the event upon ingestion. With real-time processing, Hazelcast Jet is a reliable solution for serving machine learning models that require the latest information to inform decisions.

Furthermore, Hazelcast Jet integrates with TensorFlow to run real-time classification and prediction workloads at scale. Customers can choose whether they want to use the embedded, in-process Java runner or a remote TensorFlow Serving option.

In-Memory Computing Platform

Combining Hazelcast Jet with Hazelcast IMDG enables enterprises to deploy a high-performance and scalable in-memory computing platform that handles data in motion and at rest.

Hazelcast uniquely presents customers with the ability to utilize a common architecture and skill sets to ingest and process streaming data, while also providing storage and computation of data, all at industry-leading low latency.

Availability and Resources

Hazelcast Jet 3.0 is available today for download.