Depend: a Design Environment for Prediction and Evaluation of System Dependability

This paper describes the development of DEPEND, an integrated simulation environment for the design and dependability analysis of fault-tolerant systems. DEPEND models both hardware and software components at a functional level, and allows automatic failure injection to assess system performance and reliability. It relieves the user of the work needed to inject failures, maintain statistics and output reports. The automatic failure injection scheme is geared toward evaluating a system under high stress (workload) conditions. The failures which are injected can affect both hardware and software cornportents. To illustrate the capability of the simulator, a distributed system which employs a predictionbased, dynamic load-balancing heuristic is evaluated. Experiments are conducted to determine the impact of failures on system performance and, to identify the failures to which the system is especially susceptible.