Summary: | 博士 === 國立清華大學 === 資訊工程學系 === 96 === Real-time scheduling for task sets has been studied, and the corresponding schedulability analysis has been developed. Real-time systems using rate monotonic (RM) fixed priority scheduling can be checked for schedulability either by sufficient but pessimistic schedulability conditions or by exact test. In this thesis, a series of period-dependent schedulability test methods for real-time tasks scheduled by the RM policy on uniprocessor systems are proposed.
In the first portion, we find the efficient schedulability test methods for periodic tasks. Both exact schedulability test methods and sufficient but pessimistic schedulability test methods are investigated. Exact testing provides a more precise result but may not be performed in polynomial time. Audsley et al. proposed one of the earliest methods by iteratively deriving the response times of jobs. Other researchers have improved the exact test method by using different initial values for testing. In this thesis, we first improve the iterative formula used to calculate the response times of jobs. Then, we propose a new initial value for Audsley’s test method to improve the efficiency. We show that the new initial values can significantly improve the efficiency of exact testing. In addition, a better RM schedulability bound for performing the sufficient but pessimistic schedulability is also proposed. The period ratios are considered in the schedulability bound. When more period ratios are considered, the schedulability bound can be higher.
In the second portion, we find the efficient schedulability test methods for multiframe tasks. We explore both exact schedulability test methods and sufficient but pessimistic schedulability test methods. For the exact schedulability test, the period-dependent initial value is used for the test and effectively reduce the number of iterations for testing. For the sufficient but pessimistic schedulability test, we use the information of the relative period ratios between tasks in a system to derive a new schedulability condition. By considering the smallest and the largest period values in a system, we can show that the RM schedulability bound can be improved significantly. This method also can be applied to other test methods studied earlier to improve the schedulability of real-time multiframe systems.
|