Implementations of Space-Time Correlations of Atmospheric Turbulence Using a Low-Cost, High-Performance Parallel Approach

Abstract

Modern meteorological experiments are capable of generating such large quantities of data that the processing of the results can be prohibitively expensive on a mainframe or far too time-consuming on most reasonably priced desktop systems. This paper describes a cost-effective alternative based upon the transputer, a powerful 32-bit microprocessor designed specifically for developing multiprocessor systems. A single processing node, consisting of one transputer and one megabyte of RAM, can be assembled from fewer than a dozen chips for around $1200, or purchased commercially for about $1500. The system described here consists of three such nodes hosted by a PC/AT clone and calculates the space-time correlation functions of atmospheric turbulence data about 10% faster than a VAX 8650; the system can easily be expanded to tens of nodes with a nearly proportionate increase in computational power. This multiprocessor system is as easy to program as a single processor system.