Resource scheduling in cloud computing environments faces numerous challenges—controlling task execution latency, managing total system energy consumption, and balancing load among system nodes. This paper proposes an improved multi-objective resource scheduling method based on the NSGA-III algorithm. This method constructs a multi-objective optimization framework with latency, energy consumption, and load balancing as optimization objectives. The multi-objective method employs adaptive crossover and mutation operators to dynamically adjust the probability of genetic operations, ensuring a balance between global search/exploration and local development potential. Furthermore, the method provides a resource utilization prediction initialization model, utilizing historical load data to improve the quality of the initial population, and incorporates dynamic virtual machine consolidation and migration strategies to account for system load changes. With a task workload of 10,000 tasks, the proposed method achieves an average task latency of 1420 milliseconds, reduces total system energy consumption to 765 kWh, and maintains a low standard deviation in load balancing, thus demonstrating the effectiveness and stability of the proposed method within a multi-objective collaborative optimization framework.
Wang et al. (Thu,) studied this question.