Distributed Structural Analysis of Large–Scale Structures on a Cluster of Personal Computers

Abstract

Recent developments in high–performance computing technologies have had a significant impact on structural analysis and design. However, most algorithms or computational strategies have been implemented on supercomputers and powerful workstations, which are beyond the reach of small organizations. In this paper, distributed computational models are presented for structural analysis of large–scale structures on a cluster of PCs based on two different levels of substructuring techniques. They are implemented on a collection of Pentium processors connected via a 10 Mbit/s Ethernet LAN. The effect of different levels and sizes of substructures on the performance of distributed implementation is investigated. The effectiveness of the algorithms is assessed by applying them to static analyses of two moment–resisting steel frames and three space truss structures. The algorithm based on multilevel substructuring allows a larger granularity of the parallel tasks, which can result in improved performance.