Adaptive scheduling: overload scheduling for mission critical systems

Mission critical systems are expected to perform a subset of critical functions under fault conditions. The occurrence of faults can result in shared resources, such as processors or networks, being overloaded. The goal of this effort is to provide a benefit based scheduling approach that facilitates operations under overload conditions. A key feature of this approach is the ability to detect the occurrence of faults prior to an overload by leveraging unused capacity. A benefit based scheduler that considers importance, system state, and timeliness is used to order processing once faults have been detected. This approach can be easily fitted to classical deadline based scheduling policies, such as rate monotonic or earliest deadline first, and causes no significant increase in algorithm complexity. Simulations show that the ability to order processing based on a benefit model offers considerable advantages over classical deadline based approaches during overload conditions.