ADAPTIVE CPU SCHEDULING FOR MITIGATING RESOURCE STARVATION IN MULTI-TENANT OPERATING SYSTEMS

Abstract

Multi-tenant operating systems are widely employed in cloud computing, virtualization platforms, and shared computing infrastructures to maximize hardware utilization while supporting multiple users and applications concurrently. However, resource starvation remains a significant challenge when CPU-intensive workloads monopolize processor time, leading to degraded performance and unfair allocation for lower-priority or latency-sensitive tasks. This study proposes an adaptive CPU scheduling approach designed to minimize resource starvation while maintaining system throughput and responsiveness. The proposed scheduler dynamically adjusts process priorities and CPU time allocation based on workload characteristics, execution history, waiting time, and resource utilization metrics. Extensive simulation experiments were conducted using heterogeneous workloads representing compute-intensive, interactive, and background processes. Performance evaluation considered average waiting time, turnaround time, CPU utilization, response time, fairness index, and starvation occurrence rate. Experimental results demonstrate that the adaptive scheduling strategy significantly reduces starvation incidents while improving fairness among tenants without introducing excessive scheduling overhead. Compared with conventional scheduling algorithms, the proposed method achieves lower response latency, more balanced CPU distribution, and improved overall system efficiency under varying workload conditions. The findings suggest that adaptive CPU scheduling offers a practical solution for enhancing quality of service and resource fairness in modern multi-tenant operating systems deployed in cloud and virtualized environments.