Company›Newsroom›Hazelcast Announces In-Memory MapReduce API

By Hazelcast Comms

Feb 13, 2014

Back to Newsroom

Hazelcast Announces In-Memory MapReduce API

Leading Open Source In-Memory Data Grid Gains Key Capability for Processing Big Data

For immediate release: Santa Clara, CA – At the Strata conference, Hazelcast® (https://hazelcast.com) the maker of the leading open source in-memory data grid, announces availability of the new Hazelcast MapReduce API.

In-Memory MapReduce can be up hundreds of times faster than running it on a disk farm as is traditional for systems like Hadoop. This can be a game-changer for real-time and low-latency applications, stream processing and machine learning. Hazelcast forms a dynamically and elastically scalable cluster of Java Virtual Machines (JVMs) that can create very large clusters of RAM for processing data.

Hazelcast already provides APIs for distributed computing including an Executor Service that sends workload across a cluster and Entry Processor which executes computation targeted to nodes with specific data on them. MapReduce provides another powerful paradigm for data-oriented distributed computing.

“Hazelcast does for transactional systems what Hadoop does for data warehousing.” said Talip Ozturk, CEO of Hazelcast “It uses open source to aggregate commodity hardware and in doing so, democratizes transaction processing, previously the sole domain of high end mainframe systems and the select few who know how they work. MapReduce provides a new way to harness the compute power of these distributed processors.”

“Developers often look at Hazelcast as an In-Memory NoSQL datastore.” said Christoph Engelbert, creator of the Hazelcast MapReduce API “But with MapReduce, Executor Service and Continuous Query, I would more describe it as a distributed computing environment that goes way beyond traditional NoSQL solutions.”

Hazelcast MapReduce will be combined with Hazelcast Continuous Query into Continuous MapReduce in an upcoming versions, distributed processing can be executed on streaming data directly on the wire, just like Twitter does with your tweets. Stream processing can be used for risk, security and fraud detection, for example, by applying complex algorithms on data as it flows across transactional systems. Hazelcast MapReduce leverages the inherent parallelism of a distributed cluster that enables it to scale elastically and linearly.

Links