A Comprehensive Multi-Criteria Evaluation Framework for CPU Scheduling Algorithms

Abstract

Evaluating CPU scheduling algorithms is inherently challenging due to the need to balance conflicting criteria such as efficiency, fairness, responsiveness, and starvation avoidance. However, most existing studies rely on a limited set of metrics, often leading to incomplete and potentially misleading conclusions. This paper proposes a comprehensive multi-criteria evaluation framework that enables a more balanced and reliable assessment of scheduling performance. The framework integrates key performance metrics, including average waiting time, turnaround time, response time, fairness, priority adherence, and starvation. A Simple Additive Weighting (SAW) method is employed to aggregate these metrics into a unified evaluation model. To enhance reliability, the non-parametric Friedman test is applied as a statistical validation procedure. Experimental results reveal significant trade-offs among algorithms, underscoring the limitations of single-metric evaluation and demonstrating the effectiveness of the proposed evaluation framework for systematic and informed comparison.