The desire for low-cost reliable computing is increasing. Most current fault tolerant computing solutions are not very flexible, i.e., they cannot adapt to reliability requirements of newly emerging applications in business, commerce, and manufacturing. It is important that users have a flexible, reliable platform to support both critical and noncritical applications. Chameleon, under development at the Center for Reliable and High-Performance Computing at the University of Illinois, is a software framework. for supporting cost-effective adaptable networked fault tolerant service. This thesis details a simulation of fault injection, detection, and recovery in Chameleon. The simulation was written in C++ using the DEPEND simulation library. The results obtained from the simulation included the amount of overhead incurred by the fault detection and recovery mechanisms supported by Chameleon. In addition, information about fault scenarios from which Chameleon cannot recover was gained. The results of the simulation showed that both critical and noncritical applications can be executed in the Chameleon environment with a fairly small amount of overhead. No single point of failure from which Chameleon could not recover was found. Chameleon was also found to be capable of recovering from several multiple failure scenarios.
