GridGain

GridGain is a Java-based open source grid computing framework. It is free and is dual licensed under LGPL and Apache 2.0 licenses. GridGain is based on MapReduce paradigm and allows to distribute Java code across multiple grid nodes. The MapReduce paradigm is at the core of what GridGain does. It defines the process of splitting an initial task into multiple sub-tasks, executing these sub-tasks in parallel and aggregating (reducing) results back to one final result.

Some of GridGain features include transparent grid enabling based on AOP (Aspect Oriented Programming), peer class loading, pluggable SPI-based architecture. GridGain handles automatic node discovery, automatic deployment and redeployment of your code on the grid, node topology and fail-over resolution, load balancing, affinity job mapping for data grid integration, and more.

GridGain was founded by Nikita Ivanov and Dmitriy Setrakyan in 2005 and had it first release in July 2007. The company is located in Pleasanton, CA with R&D office in Saint Petersburg, Russia.

Service Provider Interface (SPI)
GridGain has two sets of API. One is public used for grid task execution, message exchange, event querying, etc. Another is a service provider interface (SPI) layer which is used for plugging various low-level functions of the grid, such as discovery and communication, failover, topology resolution, load balancing and more. SPI Architecture allows GridGain to integrate with various open source and commercial messaging and clustering products, such as JBoss HA, MuleSource, JMS, JGroups, Oracle Coherence, Email, and others.

Distributed Unit Tests
GridGain provides ability to distribute long running JUnit tests across the grid. By simply substituting JUnit TestSuite with GridJunitTestSuite you are able to run all the tests in that suite in parallel on multiple nodes which can significantly speed up the overall time of completion.

GridGain on Amazon EC2
In the largest scalability test to date on the Amazon Elastic Compute Cloud, GridGain was used to run a Monte Carlo Simulation using from 2 to 512 nodes. The results demonstrated linear scalability across the range of nodes.

Main Menu

See Also

Main Menu