Studying Sabotage-Tolerance Mechanisms through Web-based Parallel Parametric Analysis and Monte Carlo Simulation

摘要

In this paper, we show how we have been able to use parallel parametric analysis and Monte Carlo simulations, running on a Java applet-based volunteer computing system, Bayanihan, to develop and study new mechanisms for addressing the problem of sabotage by malicious volunteers in volunteer computing and Internet computing systems. We begin by describing the general-purpose framework we have developed for writing various parametric analysis and Monte Carlo applications on Bayanihan. Then, we give an overview of the sabotage-tolerance mechanisms we have developed, including voting, spot-checking, and credibility-based fault-tolerance, and present some examples of results we have gotten using our simulator. Finally, we present performance results that show how parallelizing our simulator has made our research possible by enabling us to do parametric analysis in much less time than possible with a single computer.