Lattice Boltzmann method on a cluster of IBM RISC System/6000 workstations

There is a wide consensus among researchers that parallel computing is the most promising way to cope with the ever increasing quest of computational power raised by modern research in science and engineering. The specific way how parallelism is effectively implemented at the architectural level, remains, however still an open issue. An instance of parallel computing which is gaining a growing popularity is the one consisting of distributed environments built out of pools of networked workstations. The main merit of this approach, hereafter referred to as to `cluster computing´, lies in its cost effectiveness with respect to `traditional´ shared-memory supercomputers. A cost-effectiveness which is basically driven by the impressive development of fast RISC (reduced instruction set computer) microprocessors. An implementation of the lattice Boltzmann method on a homogeneous cluster of IBM RISC System/6000 superscalar workstations is presented