A Parametric Study of an Open-Source Distributed Computing Environment for Tribological Studies

Abstract

A parametric study of a single system image cluster is presented for parallel computing applications in the tribology field. The trend to use a cluster to solve computationally intensive tasks is emerging for high performance/cost ratio, scalability, and freely available open source. At the present time such a system is very easy to construct and to maintain for cluster computing. An air-bearing model with four design variables was used for simulating light to moderate and variable computational loads. The factors considered in this parametric study are network connecting speed, master node capability, and computing task size. The subjects of the task allocation method and task load balancing are also discussed to elaborate the characteristics of the cluster. Although computational problems in the tribology field are diversified in every aspect, the parametric survey can assist in the construction of a cluster as well as the selection of suitable parallel algorithms for computationally intensive simulations.