An Energy- and SLA-Aware Heuristic for Efficient Virtual Machine Migration and Resource Scheduling in Cloud Computing

Abstract

Cloud computing provides elastic access to virtualized computing resources, but efficient scheduling remains difficult because workloads, resource availability, and service requirements change continuously. Conventional scheduling methods can reduce execution time for selected workloads, yet they may not simultaneously address resource utilization, service-level agreement (SLA) compliance, virtual machine (VM) migration, and energy consumption. This study develops and evaluates a heuristic resource-management approach that integrates host overload/underload detection, VM selection based on migration time, controlled VM migration, and host shutdown. The proposed approach uses CPU utilization as a principal indicator of host state and applies a migration-oriented decision process to reduce unnecessary movement while maintaining service continuity. The research is grounded in a CloudSim-based simulation framework and compares the proposed policy with established threshold- and median-absolute-deviation-based VM selection and migration policies. Results reported in the thesis show that the proposed approach achieved an overall SLA violation of 27.5%, compared with 38.2% for the referenced existing approach, representing a relative reduction of approximately 28.0%. The proposed method also recorded 16 SLA-related host shutdown events in the detailed experiment, while the compared methods generally recorded 17 or more, and the thesis reports lower energy consumption than the evaluated heuristic policies. A separate baseline experiment in the thesis also confirms that Min-Min provides the shortest execution time among FCFS, SJF, Round Robin, Max-Min, and Min-Min under the stated CloudSim configuration. The findings indicate that scheduling decisions should be multi-objective rather than based on execution time alone. The proposed heuristic is therefore positioned as a practical, simulation-validated direction for energy-aware and SLA-conscious cloud resource management.