Scheduling multimedia applications under overload and non-deterministic conditions

Scheduling multimedia applications is not straightforward. The deterministic techniques in hard real-time scheduling, though applicable, are not absolutely necessary and instead result in very conservative execution. Current real-time scheduling research does not properly address the problem of scheduling under overload and nondeterministic situations. In this paper, we introduce a more representative task model for multimedia applications. We describe the use of a frequency parameter, associated with the tasks, which can serve to regulate execution during degradation (when the system is overloaded). We also analyze the worst case CPU bounds for scheduling using frequency and show the possibility of utilization, /spl Sigma/c/p>1 under certain conditions. We conduct experiments to see the performance of the model and best-effort frequency-based scheduling under extreme and variable conditions, where modeling fails.