Computing Clusters
Use multiple computers to process heavy data loads.

By Leo Laporte and Roman Loyola


High-end graphics and animation programs often use distributive processing clusters to handle heavy data loads. In a distributive-processing cluster, CPU-intensive tasks are divided over several networked computers. In essence, multiple computers are doing the job of a single computer, and that means shorter times to finish a task.

On today's show, Matt Olander of Offmyserver shows how to create a processing cluster using old computers. The cluster Olander assembles is called a Beowulf cluster. Olander also talks about using FreeBSD for clusters. Take a look at the Offmyserver cluster built for today's show.


OS support


Linux offers operating system level support for linking independent PCs. This technology is called "clustering." It's usually implemented for one of three reasons: parallel processing, load balancing, or fault tolerance.


Clustering is supported in Windows Server 2003.


Mac OS X Server supports clustering. Read how Carnegie Mellon University implemented its Mac OS X clusters.


Cluster links


Learn more about clusters by clicking on the links below.



FreeBSD cluster forum
Mailing list for technical discussions related to creating clusters of FreeBSD computers.

Beowulf.org
Comprehensive resource for Linux clustering.

Beowulf cluster mini-how-to
Written by Ole H. Nielsen at the Technical University of Denmark

Guide to Creating and Configuring a Server Cluster under Windows Server 2003
White paper with step-by-step instructions for creating and configuring a typical single quorum device multi-node server cluster.

Deploying Mac OS X in Clusters
Carnegie Mellon University's deployment plan.

PS2 Supercomputer
How PlayStation 2 devices are linked together to create a cluster.


Originally posted December 17, 2003