DETERMINING REAL-TIME PROCESSING RESERVE IN MULTITASKING SYSTEMS

The ability to accurately predict usable processing reserve is needed in many real-time software applications. Earlier research has shown that there are limits on the amount of available processor throughput that can be used to perform work with hard real-time deadlines. In this paper, we present an approach for determining and manipulating usable real-time processing reserve for multitasking systems that use preemptive, priority-based scheduling rules. Such systems must be prioritized using either rate monotonic, or deadline monotonic criteria. The approach computes real-time processing reserve for a schedulable task set, assuming that additional computation requirements will be levied on existing tasks. The approach includes a method for computing the maximum work that can be performed by an added task while preserving schedulability of the task set. Additionally, we introduce a method for balancing work within a task set to increase available real-time processing reserve. The method can be conditionally applied to harmonic tasks within a task set to improve the work efficiency of a processing resource.