Thread: High Performance Clustering

Open Source can be called the father of commodity HPCs (high performance clustering). HPCs got into mainstream application due to Open Source move¬ment and penetration of Linux worldwide, which made commodity HPCs a reality. And why this happened, can be understood very easily.

In the eighties and early nineties the only way to deploy a Supercomputer or an HPC was to go to some highly special ized companies like Cray, etc, then acquire a good bunch of specialized hardware and software equip¬ments and get them assembled. But owing to the requirement of specialized hardware and software and the market niche, the cost of the whole setup used to run very high, and it was not feasible for any standard organization to own such a setup. Thus only big educational institutes and government research cen¬ters could afford such HPCs.

But, in the mid nineties the concept of commod¬ity HPCs came into picture. And to build a commodity HPC only two things were required. First is of course the commodity hardware by which we mean the standard PC and hardware equipments. And the second thing is an HPC middleware which could be commodity. This second requirement was very efficiently addressed by Open Source developers and in no time we saw a plethora of Open Source HPC/SSI/Grid middleware available. Some such noteworthy examples could be OSCAR, OpenMosix, ROCKS, Globus, etc.