Megacomputer
From Wikipedia, the free encyclopedia
This article may require cleanup to meet Wikipedia's quality standards. Please improve this article if you can. (January 2007) |
A megacomputer is the term used to describe a group or cluster of interconnected supercomputers.
[edit] Overview
Megacomputers are the next level in the hierarchy of parallel computing. This hierarchy can be compared to that of the universe. Each individual component (node) is a part of a larger group of the same components (cluster), and that cluster becomes a node of a larger group. A megacomputer is a cluster of supercomputers.
[edit] First Megacomputer
The first known megacomputer was the result of a European project called DEISA (Distributed European Infrastructure for Supercomputing Applications), where eight European research institutes linked four supercomputers to create a megacomputer to boost Europe's competitiveness in scientific areas where extreme performance is needed. Below is a description of the current status of DEISA as of 26th November 2006. [1]
The DEISA network DEISA uses an internal network provided by GEANT and the National research networks (NRNs) that connects the supercomputers (not the sites) and offers reserved bandwidth, This internal network exists, of course, in addition to the standard Internet connectivity that each national supercomputer centre offers. Today, in the first phase of the project, the internal DEISA network provides 1 Gb/s connectivity between the four IBM supercomputers in Germany, France and Italy integrating the initial AIX super-cluster. The purpose of the phase 1 network is to enable the operation of the AIX super-cluster from May 2005. In the short term in 2005 the DEISA internal network will evolve by connecting the IBM platform in Finland to the AIX super-cluster, and by providing dedicated connectivity to SARA’s Linux system. In the medium range term, DEISA – in collaboration with GEANT and the NRENs - is beginning the design of a phase 2 network, based on the new generation European network infrastructures. Reserved bandwidth multi-gigabit connexions are planned not only inside the AIX super-cluster, but also across all supercomputing resources in the extended, heterogeneous grid.