SourceForge - Introduction to Single System Image Clustering
Thursday December 20 2001 @ 08:25AM EST

"The purpose of this article is to introduce Single System Image (SSI) Clustering. First we will look at what clustering is and the various problems a clustered platform solves. Then we generally define what a SSI cluster is and look at various forms and degrees of SSI. From there we focus on the Open SSI Cluster Project - the goals, the benefits SSI clustering can bring and the application areas for SSI clustering. We complete the paper with an architectural overview of the SSI cluster project.

The term clustering, with no adjective specifying the type of clustering, means different things to different groups. Below is a short discussion of high performance (HP) clusters, load-leveling clusters, web-service clusters, storage clusters, database clusters and high-availability (HA) clusters. All of these have a few things in common that distinguish clusters from other computing platforms.

Clusters are typically constructed from standard computers, or nodes without any shared physical memory. That, combined with the characteristic of running an OS kernel on each node, distinguishes a cluster from a NUMA (Non-Uniform Memory Architecture) or SMP (Symmetric Multi-Processor) platform (of course a cluster could be constructed from NUMA or SMP nodes). The "purpose" of a cluster is for the individual nodes to work together to better provide HA (High Availability), HP (High Performance), parallel processing or parallel web servicing. Finally, a cluster is different from a client-server distributed computing environment in that each of the cluster platform solutions is more peer-to-peer, typically with a sense that each node is a member of a larger system."


< Sistina - Global File Systemâ„¢ Version 5.0 | ZDNet - Server blades: Density versus flexibility >