Resource Reclaiming in Multiprocessor Real-Time Systems
IEEE Transactions on Parallel and Distributed Systems
| Hi-index | 0.00 |
TR: 90-89 TR Title: Resource Reclaiming in Real-Time Authors: C. Shen, K. Ramamritham, J.A. Stankovic Address: Computer Science Department, LGRC University of Massachusetts Box 34610 Amherst, MA 01003-4610 Date: February 11, 1991 Most real-time scheduling algorithms schedule tasks with respect to their worst case computa- tion times. Resource reclaiming refers to the problem of utilizing the resources left unused by a task when it executes less than its worst case computation time, or when a task is deleted from the current schedule. Resource reclaiming is a very important issue in dynamic real-time multiprocessor environments. Resource reclaiming algorithms must be designed with three issues in mind. Firstly, a resource reclaiming algorithm should be effective in reclaiming resources; that is, it should improve the performance of the system. Secondly, since most of the tasks execute less than their worst case computation times, resource reclaiming will occur frequently. To be effective, resource reclaiming overheads should be small. More precisely, we seek an algorithm with bounded overhead, i.e., with time complexity that is independent of the number of tasks in a schedule, and include this overhead cost in the worst case computation time of a task. Thirdly, when the actual computation time of a task differs from its worst case computation time in a multiprocessor schedule with resource constraints, run time anomalies may occur if greedy schemes are used for utilizing idle resources. These anomalies may cause some of the already feasibly scheduled tasks to miss their deadlines. Thus, dynamic resource reclaiming algorithms must be effective in reclaiming unused time, have low, bounded overheads, and also avoid any run time anomalies. In this paper, we present resource reclaiming algorithms with these properties. The effectiveness of the algorithms is demonstrated through simulation studies. The algorithms have also been implemented in the Spring Kernel [17]. index terms --- deadlines, dynamic real-time systems, multiprocessor scheduling, resource con- straints, worst case computation times.