Scheduling Strategy to Minimize Makespan for Energy-Efficient Parallel Applications in Heterogeneous Computing Systems

Energy consumption has emerged as a critical design constraint in heterogeneous computing systems, spanning from small embedded devices to expansive data centers. In this paper, our primary focus is on the challenge of minimizing scheduling lengths for parallel applications within energy-constrained heterogeneous computing environments. Here, the scheduling length denotes the actual time required for a task to reach completion. In this study, we tackle the issue of minimizing energy allocation for unassigned tasks and introduce a novel task scheduling algorithm (EEMM). This algorithm incorporates a weight-based mechanism for pre-assigning energy consumption to unassigned tasks. Through a series of experiments conducted on real parallel applications, we consistently observe that the proposed algorithm ensures that the actual energy consumption remains within specified constraints and achieves shorter scheduling lengths. This demonstrates its superior performance. This research offers a valuable solution to the task scheduling problem in energy-constrained heterogeneous computing environments.