Energy-aware scheduling approach for enhanced performance of parallel applications in simulated heterogeneous cloud environments

Energy-constrained workflow scheduling seeks to minimize application makespan while preserving a finite energy budget across heterogeneous DVFS-enabled resources. Existing pre-assignment methods generally distribute energy using a single task-level criterion, which can restrict the VM–frequency choices of later tasks when the budget is tight. This paper proposes the Energy-Aware Scheduling Approach (EASA), whose core scheduling procedure combines four elements: a normalized dual-factor surplus-energy allocation based on both minimum computation cost and minimum energy demand; explicit reservation of pre-assigned energy for unscheduled tasks; dynamic derivation of the energy budget available to the current task; and systematic evaluation of feasible VM–frequency candidates using earliest finish time. Theoretical analysis proves that the pre-assignment preserves the application-level energy budget and that the generated schedule satisfies the prescribed energy constraint. EASA is evaluated through controlled simulation on six benchmark scientific workflows—Gaussian Elimination, Linear Algebra, Fast Fourier Transformation, Genome, LIGO, and Montage—over multiple workflow sizes and energy-budget settings. The results, supported by sensitivity analysis and paired non-parametric statistical tests, show that EASA obtains the best overall average rank and its strongest makespan advantage under strict energy constraints. The study is limited to offline simulation-based evaluation; real-cloud measurement and online adaptation are left for future work.