Originally Written by Enes Akar.

I am excited to announce the EA (Early Access) release of the upcoming Hazelcast 3.5 version. After ending 2014 with a fully JCache compliant Hazelcast 3.4 version, we have focused on increasing stability and performance in this next release as well as solved many of the open bugfixes for 3.5.

On the stability side, we have added an Async Back Pressure system to prevent the system from overloading, a feature that you will see further development and optimization of in the near future. In addition we just recently released Hazelcast Simulator – used to test Hazelcast itself but now available to test your own applications.

Also, scalability of MapLoader and CacheLoader was enhanced to quickly load large data sets from external resources. Additionally MapReduce performance, when used with KeyPredicate or a set of keys, as well as query speed on IMap and ReplicatedMap were improved radically.

Obviously that’s not all yet. Even though these enhancements are important to current users, our restless engineering team around the world added new, major features as well. After almost 4 months of development we are happy to ask early adopters, from both the community and enterprise customer side, to test them and tell us how it went and what they think.

New Hazelcast Open Source Features Added Cluster Quorum which blocks the cluster whenever the number of nodes drops under a certain threshold to prevent inconsistency on network failures.

Another amazing new feature is the RingBuffer implementation. This new data structure implements a circular buffer with fixed capacity for low latency access and fixed resource consumption. Built on top, we now offer a reliable ITopic system with guaranteed delivery.

Last but not least we broke down our EntryListener type into several interfaces, each interface for a single type of event. This approach adds more event interfaces without breaking the users’ code when upgrading (Open/Close Principle), and adds a new PartitionLostListener to listen for events when data partitions get lost in failure situations. One of the top requests from the community.

Further enhancements include network level and protocol optimizations, a ConfigPatternMatcher to match wildcard rules, support for Amazon v4 Signatures (AWS Joiner), optimized XA Transactions and Import statement for Hazelcast client configurations.

New Hazelcast Enterprise Features Added: With those open source updates in mind, lets have a look at some of the additions that went into Hazelcast Enterprise:

Continuous Query Caching offers the possibility to pre-populate a Map and having it updated automatically using a given query. WAN Replication now guarantees order of updates and optimized network traffic by batching replication events. Also new High-Density backed Near Cache provides a local storage to keep recently or hot entries next to the consumer. Using this technique network roundtrips are avoided and read operations are dramatically speeded up.

And last but not least, Hazelcast 3.5 will include a hugely improved new client protocol. Hazelcast 3.5 will offer a first public preview of this new client protocol. A data protocol to support backwards and upwards compatibility to offer the option of seamlessly updating a cluster from one version to another. The protocol is fully specified and open to anyone to implement with external clients compatible with Hazelcast. I am looking forward to seeing the new client code being tested and hearing your opinions and experience.

The Hazelcast team hopes that you find these additions and optimizations as exciting as we do. As early adopters, I invite you to try it out and share your comments with the community. Download Hazelcast 3.5 EA today or use your favorite dependency manager with the following

<dependency><groupId>com.hazelcast</groupId>
<artifactId>hazelcast</articaftId>
<version>3.5-EA</version>
</dependency>