Comparing Modern Differential Evolution Variants in Multiobjective Cloud Scheduling

Cloud scheduling is a complex optimization problem balancing fast task execution for cloud users and efficient resource utilization for cloud operators. The requirements by both parties are usually limited by Service Level Agreements and Quality of Service constraints that define the level of provided service. Although many scheduling algorithms exist, the ability of nature-inspired metaheuristics to solve complex multiobjective optimization problems motivates their investigation in the context of cloud scheduling. This work evaluates the ability of modern Differential Evolution algorithm versions to schedule cloud jobs. The computational experiments are performed with a recent cloud scheduling dataset and demonstrate a large variance in the results obtained by different algorithms. Finally, is pointed out that cloud scheduling is a real-world constrained multiobjective optimization problem with a good potential to serve as a benchmark for nature-inspired metaheuristics.