Dashboard > GridGain User Guide > Table Of Contents > Developers Guide > GridGain Architecture
GridGain Architecture
Added by morpheus, last edited by morpheus on Jun 30, 2008  (view change)
Labels: 
(None)


Public API

At the top of GridGain lies public Grid API with which user interacts. Via this API user has ability to execute tasks, get grid-specific information, such as topology or events, send direct messages to remote nodes, etc..

Grid Kernal

Below public API lies Grid Kernal. Note the misspelling, Kernal instead of Kernel. We named the core of GridGain Kernal in honor of the old Commodore's operating system in its 8-bit home computers. See Wikipedia for more details. The responsibility of Kernal is to provide implementation for GridGain public API and delegate necessary functionality to SPI Layer.

Service Provider Interfaces

SPI (Service Provider Interface) Layer is a set of Java interfaces with various implementations for them. GridGain comes with many SPI implementations out of the box. SPIs are used to provide various implementations for node discovery, communication, topology resolution, job checkpoints, scheduling and job collision resolution, load balancing, fail-over, etc... SPI's are used by GridGain internally and are not open for public use. They enable GridGain to easily switch, say, between Multicast-based discovery and JGroups-based discovery just by changing a couple of lines in configuration file.

The diagram below visually depicts GridGain architectural layers and outlines the SPI implementations shipped out of the box. The SPI's colored in gray are the ones started by default without any changes to GridGain configuration.

Powered by Atlassian Confluence, the Enterprise Wiki. (Version: 2.2.10 Build:#528 Nov 29, 2006) - Bug/feature request - Contact Administrators