Types of Grid
This article uses the term "Grid" in a loose sense, meaning a specific computing infrastructure designed to aggregate a set of resources. Using this definition, both the SETI@home project and a Beowulf cluster could be considered Grids.
This article tries to define the different types of Grid which are in common usage. Grids can be broadly categorised by resource, scale and service.
Grids by Resource
Grids can be classified by the type of resource they share. Some Grids may share more than one type of resource, and therefore fall into more than one of the following categories:
- Computational Grids. These primarily share CPU resources. Examples include TeraGrid which has more than 750 teraflops and SETI@home which aggregates the computing power of millions of home computers. Utility computing ventures such as the Sun Grid Compute Utility are also an example of computational Grids.
- Data Grids. These are intended to share data resources, such as the results of experiments, between users. Typically, data Grids will handle very large amounts of distributed data, such as QCDGrid or the LHC Computing Grid.
- Storage Grids. These are designed to provide users with access to an enormous amount of storage space. One of the biggest and most well-known examples of a storage Grid is Amazon S3.
- Equipment Grids. Grids can also be set up to share access to physical resources, such as astronomical telescopes in the eSTAR project.
Grids by Distribution
Grids can also be classified by how geographically distributed their resources are:
- Internet. Internet-scale Grids can potentially include anyone with access to the internet. Examples include SETI@home and the World Community Grid which both contain thousands of computers dispersed across the globe.
- Virtual Organisation (VO). A VO-scale Grid contains several academic or corporate entities. Most Grids fall into this category e.g. TeraGrid, QCDGrid.
- Local. A local Grid is contained entirely within one organisation. Often companies will have access to their own clusters for processing tasks. For example, the render farms used to produce animated films such as Toy Story are a form of local Grid.
Grids by Service
From a user perspective, the important thing about a Grid is the services it provides. For example, a render farm will provide graphical rendering services using a specific software package such as Blender or Shake. A data Grid gives access to specific data resources, such as results from a large physics experiment.
Although many computational Grids are application agnostic, some Grids may not be able to run certain applications due to licensing or platform restrictions. Therefore, as Grid becomes more prevalent, we may start to see more initiatives such as Network.com which provides a catalogue of applications that can be run on the Sun Grid computing facility.
"Cloud Computing" is often categorised by application. Cloud Computing generally refers to users off-loading application processing to some 3rd party resource or "cloud". Normally the "cloud" can be defined as some form of Grid, often using virtualization technology to optimize and simplify resource usage. Cloud computing can be broken down by the services it offers to consumers, typically web applications such as Google Apps.
At the moment, the most prevalent types of service offered by Grids are forms of graphical rendering, scientific simulations and web applications.
Further Information
Another useful categorisation of Grids can be found at the Grid Café. IT-Tude.com also hosts a domain-based categorisation of Grids from the BEinGRID project.
Read on ...
Grid Infrastructures - examples of high profile Grids from business and research.
Case studies - showing these types of Grid put to real use.









