BEOWULF CLUSTERS

In the mid 1990s, high performance computers were limited to expensive proprietary supercomputers. Donald Becker and Thomas Sterling, working at The Center of Excellence in Space Data and Information Sciences (CESDIS) and sponsored by NASA Earth and Spaces Science (ESS), began research into a cluster of commodity based system as a cost effective alternative. Their basic outline architectural design of commodity based systems connected through a private network, now referred to as a Beowulf class cluster computer, is now recognised as a new category within the High Performance Computing (HPC) community. The trend to deploy HPC clusters is clearly demonstrated in the number of clusters listed in the “TOP500 Supercomputer Sites” list, which lists the 500 most powerful supercomputers in the world, and makes clustered systems the most common high performance computer architecture.

HPC clusters are typically deployed for parallel computing to aggregate more processing power or effective memory for a solution of problems within the scientific or research and development arenas. Massively parallel applications use message passing paradigm libraries, such as Message Passing Interface (MPI) or Parallel Virtual Machine (PVM), to distribute and collate data between disparate processors. The choice of private network, or interconnect, can affect the execution to of parallel applications. Interprocessor communication of large data sets require an interconnect capable of high and sustained throughput, whereas small message packets require low latency interconnect to accelerate parallel program execution

Beowulf class compute clusters will continue to play a significant role in the high performance computing arena as cost effective microprocessor technology and networks continue to evolve to ever high performance.