Integrating Best-Effort Policies into Hard Real-Time Systems based on Fixed Priority Pre-emptive Scheduling

Analysis provided for fixed priority pre-emptive scheduling enables a priori guarantees to be given regarding the timing requirements of critical tasks. By comparison, Best-Effort scheduling maximises system utility, but at best provides only probabilistic guarantees that time constraints will be met. In this paper, we show how these two schemes can be integrated. A new Adaptive Value-Density Threshold policy is introduced which approximates the Best-Effort policy but incurs much lower overheads. This policy is used to admit optional (utility enhancing) tasks for on-line acceptance testing. If guaranteed such components are assigned a fixed priority and executed when no higher priority tasks are runnable. This approach combines the benefits of guaranteeing hard requirements with increased system utility, resilience to software failures and increased application robustness.