Wuhan Ligong Daxue Xuebao (Jiaotong Kexue Yu Gongcheng Ban)/Journal of Wuhan University of Technology (Transportation Science and Engineering)

Abstract : This study conducts a comprehensive evaluation of scheduling algorithms specifically designed for independent, non-preemptive tasks within multiprocessor systems that operate under soft real-time constraints. The research introduces the Optimized Performance-Based Genetic Algorithm (OPBGA) and compares its performance against that of the Earliest Deadline First (EDF), Least Laxity First (LLF), and the Evolutionary Fuzzy-Based Scheduling Algorithm (EFSBA). The evaluation encompasses three distinct processor configurations. Utilizing randomized task sets with varying load factors, the study simulates low, medium, and high demand scenarios, ensuring a thorough performance assessment. Key performance metrics, including Average Turnaround Time (ATAT), Average Response Time (ART), and Deadline Misses (DLM), illustrate the enhanced scalability and efficiency of OPBGA. In circumstances of elevated load, traditional algorithms such as EDF and LLF exhibit significant limitations, while EFSBA demonstrates only moderate adaptability. In contrast, OPBGA consistently achieves the lowest ATAT and ART with a record of zero DLM across all configurations. The findings position OPBGA as a robust and scalable solution for real-time scheduling in multiprocessor environments, optimizing task allocation and ensuring superior performance across diverse load conditions.