Beowulf parallel processing for dynamic load-balancing

The Beowulf parallel-processing architecture is a public-domain software technology developed by NASA to Link commercial off the shelf (COTS) processors yielding super-computer performance (10 Gflops) very economically (under $150K). It has been used for applications as diverse as scalable hierarchical particle algorithms for galaxy formation and accretion astrophysics, phase transitions in the early universe and their analogues in the laboratory, and data analysis from the spacecraft in the Solar-Terrestrial Probe Line. Beowulf clusters have been ranked among the top 125 world-class supercomputers. Beowulf runs on a variety of platforms (workstations, Win-tel PCs, Macintosh) and a variety of generations of hardware. It has been successfully demonstrated for heterogeneous parallel processing with hundreds of diverse platforms. This paper describes our implementation of a Beowulf cluster comprising 18 Sun workstations and the Linux operating system. We describe and demonstrate ParaSort, a distributed, parallel, data-allocation-sorting algorithm with automatic distributed load-balancing, and fault-tolerant performance. The load-balancing feature can provide dynamic, on-board, adaptive optimal distribution of processing tasks across a heterogeneous network of computing devices. Furthermore, it can intelligently allocate processing to off-board resources as appropriate and as they become available. We analyze the strengths and weaknesses of Beowulf and describe future extensions of our work.