In-Memory Data Grid

Hazelcast 3.6 with options for cloud and container deployment

Michael Thomas
Abstract computer graphics background image via shutterstock

The Hazelcast In-Memory Data Grid implemented in Java has appeared in version 3.6. In the words of Hazelcast CEO Greg Luck, the release marks “the largest development leap of the past two years.”

Cloud Management, Container Deployment and Client Development

Hazelcast 3.6 offers new options for cloud management and container deployment which should facilitate the implementation, configuration and administration of cloud or hybrid deployments on site. Thus the Hazelcast Cloud Discovery Service Provider Interface (SPI) puts cloud-based or local nodes in a position to recognize each other with very little configuration and administrative effort respectively. The providers supplied in version 3.6 include AWS, Consul, Etcd, ApacheJClouds and Kubernetes. In addition, Hazelcast 3.6 also brings new container deployment options for Docker, including support for PaaS providers such as OpenShift and CloudFoundry.

Moreover, the availability of the Open Binary Client Protocol was announced, with the aim of a collaborative development of new clients, as well as platform independence. Until now, only clients for Python, Ruby, NodeJS, Scala and Clojure Open Source were made available, with version 3.6 however, Hazelcast’s native clients for C++ and .NET are provided as open source.

Hot Restart Store and HDMS performance features

In addition, Hot Restart Store offers users a new persistence feature. With the help of Hot Restart Store, key e-commerce and enterprise applications can start up with their data even during operation, without having to reload the data from another recording system. Here start-up time for SSD is around 1.3 GB per second per node. During operation, 310,000 write operations per second per node are supported. In the worst case scenario of a data center failure, a safety backup is on standby. Hot Restart Store supports data structures of JCache, Map, Hibernate and Web Sessions.

Last but not least, Hazelcast 3.6 comes with expanded HDMS performance features for JCache, Map, Hibernate and Web Sessions data structures. As a result, memory caches for large horizontal or vertical scaling of applications are possible: with HDMS, each node can save several hundreds of GB of data. As Hazelcast developers promise, no overhead is generated in the process through garbage collection.

Hazelcast, winner of the JAX Award for the “Most Innovative Open Tech Business” in 2014, is an In-Memory Data Grid for NoSQL databases which is based on Java and offered as open source and as an enterprise platform. Hazelcast is geared towards the optimization of NoSQL databases and falls back on distributed data structures like Map, Queue, MultiMap, Topic or Lock.

Michael Thomas
Michael Thomas studied Educational Science at the Johannes Gutenberg University in Mainz and has been working as a freelance author at since 2013.

Inline Feedbacks
View all comments