Adaptive Fault-Resistant Systems

The combined effects of faults and resource failures, wide swings in service demand, and situation-dependent user requirements, stress a computers ability to satisfy its service expectations. This is an especially significant problem in distributed system that employ unreliable communications, and whose components may operate in different and perhaps harsh physical data and usage environments. One of the goals for adaptive design is to allow flexible use of available resources to cover a much wider range of different kinds of environmental variables than could be covered by a fixed, worst-case design. The research focused on the tasks of 1 developing a theory of adaptive fault- resistant systems and general principles of architectural design 2 developing specific architectural design techniques and 3 demonstrating adaptive designs. Three mechanisms were investigated 1 Adaptive Distributed Recovery Blocks ADRBs, a multiple-mode scheme for error detection and recovery, useful for both hardware and software faults 2 adaptive fault tolerance for hybrid faults, an economical technique for tolerating both simple and complex fault types and 3 adaptive distributed thread integrity, a technique for detecting and repairing thread breaks in a wide range of operating environments using the Alpha programming model. Adaptive distributed systems, Recovery blocks, Fault tolerance, Anomaly management.