Algorithms for scheduling imprecise computations

The imprecise computation technique, which prevents timing faults and achieves graceful degradation by giving the user an approximate result of acceptable quality whenever the system cannot produce the exact result in time, is considered. Different approaches for scheduling imprecise computations in hard real-time environments are discussed. Workload models that quantify the tradeoff between result quality and computation time are reviewed. Scheduling algorithms that exploit this tradeoff are described. These include algorithms for scheduling to minimize total error, scheduling periodic jobs, and scheduling parallelizable tasks. A queuing-theoretical formulation of the imprecise scheduling problem is presented.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>