Abstract
Recently, as the rapid development of hardware technology of multiprocessor and multi-core systems, research issues related to multiprocessor real-time systems become more and more popular. Scheduling algorithm is a very important topic in this domain. It has been proved that there are optimal solutions for single processor real-time systems, such as EDF and LLA algorithm. However, on multiprocessor systems, there are more factors have to be considered and the scheduling problems are more complex than on single processor systems. In this thesis, we want to find a sub-optimal real-time scheduling algorithm for a restricted condition, such as the number of processors, the number of tasks in a task set or the task period, in the environment of multiprocessor systems. After studying RM, EDF, and LLA algorithms, and a lot of analysis for scheduling problems and algorithms, we found that EDZL has a good schedulability property. We focus on EDZL algorithm, and designed five ZL-based scheduling algorithms－REZL, UTZL, SRCZL, LRCZL and MLRCZL, based on the concept of Zero Laxity. We generated a lot of task sets randomly, and measured the average successful ratio and the average number of preemptions for these algorithms. According to the analysis of the simulation results, we make some conclusions for the performance evaluation of EDZL and other ZL-based scheduling algorithms for multiprocessor real-time systems.