Large-Scale Benchmarking of Classical, Adaptive, Hybrid, and Multi-Criteria CPU Scheduling Algorithms Across Synthetic and Real-World Workloads

Abstract

CPU scheduling is one of the major challenges in operating systems, requiring a balance among efficiency, responsiveness, overhead reduction, fairness, priority compliance, and starvation prevention. This study presents a comprehensive evaluation of classical, adaptive, hybrid, and multi-criteria CPU scheduling algorithms under unified experimental conditions. A total of 93 scheduling configurations were evaluated, including 29 established algorithms and 64 HACS-DTQSAW configurations, using a  custom simulator. Experiments covered diverse workloads and real-world datasets, with nine  performance metrics analyzed using the Friedman and Nemenyi statistical tests. Results revealed  performance variations depending on workload characteristics; SJF and Optimized SJF achieved strong  performance in selected metrics, while PRIORITY provided the best priority compliance. HACS DTQSAW demonstrated high robustness and balanced performance, highlighting the importance of multi-criteria evaluation for selecting suitable CPU scheduling algorithms.