D_EDF: An efficient scheduling algorithm for real-time multiprocessor system

A scheduling algorithm decides a schedule for a set of tasks. There are numbers of algorithm for scheduling tasks on a processor. Some of these algorithms are used for scheduling tasks on multiprocessor system either under the partitioning scheme or under the global scheduling scheme. The most common scheduling algorithms are: Earliest Deadline First (EDF) and Least Laxity First (LLF). They are optimal scheduling algorithms for single processor system, but problem arises when algorithms are used for multiprocessor system. In this paper, we have proposed a new algorithm, D_EDF. D_EDF scheduling algorithm overcomes the limitations of dynamic algorithm during overloaded conditions. The proposed algorithm D_EDF, simulated and tested for independent, preemptive, periodic tasks on tightly coupled real-time multiprocessor system under global scheduling. The performance is measured in terms of Success Ratio and Effective CPU Utilization. From experiments and result analysis it concludes that the proposed algorithm is very efficient in both underloaded and overloaded conditions. It performs always better than conventional EDF algorithm. The algorithm proposed in the paper performs quite well during overloaded conditions.