Dynamic scheduling of tasks in cloud computing applying dragonfly algorithm, biogeography-based optimization algorithm and Mexican hat wavelet

Cloud computing maps tasks to resources in a scalable fashion. The scheduling is an NP-hard problem; thus, the scheduler chooses one solution from among many. This is the reason why finding the best optimal solution, especially at a high scale of the system, is not possible. Applying metaheuristic algorithms to find a near-to-optimal solution, not the best one, could be the right approach. Dragonfly metaheuristic algorithm explores and exploits a solution space by the inspiration of hunting and emigration behavior of dragonflies in nature. But it suffers from the premature convergence of the algorithm to an undesirable when explores the solution space. In this research, an improved dragonfly algorithm (applying biogeography-based algorithm, Mexican hat wavelet and dragonfly algorithm-BMDA) is presented to resolve the premature convergence by applying a mutation phase that is the combination of the biogeography-based optimization (BBO) migration process and Mexican hat wavelet transform in dragonfly algorithm. Then, it is applied for dynamically scheduling tasks under the BMDDSF framework (BBO-Mexican hat wavelet-dragonfly dynamic scheduling framework) in the cloud computing environment. The purpose is customizing a metaheuristic algorithm to be applied in the resource manager of cloud computing to improve its performance. The BMDA algorithm was firstly evaluated for the mean error in comparison with the baseline algorithms using the CEC2017 benchmark functions. Then, the performance of the BMDDSF framework in cloud computing was tested using NASA parallel workload compared with baseline methods in terms of execution time, response time and reduction of service-level agreement violation. The experiments showed that the presented method outweighed the baseline approaches.