Hazelcast, the leading open source in-memory data grid with 500,000 installed nodes and over 16 million node starts per month, today announced the general availability of Hazelcast 3.7. The latest release is 30% faster than previous versions and is the first fully modularized version of Hazelcast. Each client/language and plugin is now available as a module – speeding up the development process for open source contributors, with new features and bug fixes released as modules alongside Hazelcast 3.7. Hazelcast 3.7 also features native Cloud Foundry integration.

The previous version of Hazelcast (3.6) was already the fastest In-Memory Data Grid (IMDG) and caching solution, as demonstrated in the following published benchmarks: Oracle Coherence Benchmark Suite, Red Hat Infinispan Benchmark, Pivotal Gemfire/Apache Geode Benchmark, GridGain/Apache Ignite Benchmark , and Redis benchmarks.  To make Hazelcast 3.7 even faster,  the networking layer was reworked for greater concurrency. The graph below illustrates the performance improvements in 3.7 (these numbers are for a four node cluster all measured on a 1Gbps network):

Cloud developers can easily drop Hazelcast into their applications and achieve the same kind of high performance scalability and in-memory performance enjoyed by investment banks. With the release of Hazelcast 3.7, Hazelcast can now work in nine cloud environments and can easily be extended via Cloud Discovery Plugins to more. For PaaS, Hazelcast is now available as a service on Cloud Foundry and OpenShift. Also, Hazelcast includes container deployment options for Docker.  Hazelcast is now also live on the Microsoft Azure Marketplace. AWS and Azure offer availability zones for higher availability. In 3.7, you can now specify a Partition Strategy of ZONE_AWARE. This allows a single cluster to run across multiple availability zones with backups kept in separate zones to primary data. An entire availability zone can be lost and the cluster keeps running.

Hazelcast continues to expand its footprint beyond its traditional Java heartland. The Hazelcast open source community has created clients for a host of programming environments including Java, Scala, .Net/C#, C++, Python, Node.js and Clojure.  Java and Scala can be used for both clients and embedded members.

In Hazelcast, you could always set an eviction policy from one of LRU or LFU. Now, with the new custom eviction policies, open source developers have more flexibility to suit custom requirements of a particular application. 3.7 provides a custom eviction policy for both Map and JCache implementations. Another enhancement is the introduction of a new plugin, Apache Spark Connector. It enables Hazelcast Maps and Caches to be used as shared RDD caches by Spark using the Spark RDD API. Both Java and Scala Spark APIs are supported. Additionally in 3.7, we introduce the DurableExecutorService which adds execute at least once fault tolerance to Hazelcast ExecutorService, a popular way to do distributed computation in Hazelcast.

Greg Luck, CEO of Hazelcast, said: “Hazelcast 3.7 is open source as its best. Our language support and cloud support has exploded. With our new modular architecture we have 29 modules out of the gate. And it makes it really easy for the community to add support for their favorite language or add a missing feature by simply creating a module for it and then contributing that to Hazelcast via our Hazelcast Enhancement Proposals program.”

To download the latest version of Hazelcast, go to www.hazelcast.com.