Clustering

A cluster can be defined as a group of two or more interconnected computers (nodes) that collectively increase application availability or performance - or a combination of those paradigms - via software and communication frameworks.

The concept of clustering itself is not new; however, the application of clustering concepts, such as high performance computing and high throughput computing, has been resurgent as of late especially in technical computing, the sciences and engineering research and development.

The business world has also started to harness the power of high performance computing clustering to arrive at answers to complex financial data modeling questions much quicker than if they were using traditional multi-processor supercomputers - which would tend to be significantly more costly to own and operate. This of course is in addition to the continued deployment and use, in the business community, of high-availability clusters which are primarily in place to ensure application availability and business continuity in the event of single and/or multiple hardware system failure(s).

Learn more

Resources

PDF_icon  All you need to know about
clustering - Part 1

PDF_icon  All you need to know about
clustering - Part 2

PDF_icon  Parallel Virtual Filesystem

Open Industry Standard Rack

Advantages of Clustering

  • A cluster can provide increased performance and scalability over a standard shared memory vertical-scaling system. The breadth of performance scalability is manifest in the ability to increase processing resources both at the node level, within individual systems, and at the cluster level - by increasing the number of individual nodes in the cluster.
  • Application availability can be increased by sharing the responsibility of serving application data across multiple nodes. If there is a requirement for running multiple applications simultaneously, they can be spread across nodes and consolidated by a resource manager in the event of single or multiple node failure(s).
  • Computational power previously available only to government level research organizations in the form of SMP supercomputers can now be harnessed by almost anybody through the implementation of clustering technologies.