With the development of science and technology,cloud computing is integrated into human life just like water and electricity,and serves all walks of life in human society.In a complex and ever-changing cloud computing environment,workflow application scheduling is an important research area.Since the mapping between workflow tasks and resources is NP-hard,the scheduling algorithm cannot find the optimal scheduling solution in a limited time.Currently,heuristic and meta-heuristic algorithms are widely used for workflow scheduling in cloud computing environments.They have better performance but require more computing resources.Although the traditional scheduling algorithm is more efficient and consumes less computing resources,the final scheduling effect is unsatisfactory.In order to solve these scheduling problems in the cloud computing environment,this thesis conducts a detailed investigation and summary of the existing meta-heuristic scheduling models and algorithms,and by analyzing the advantages and disadvantages of the cloud workflow scheduling based on multi-objective particle swarm optimization algorithm,a cloud workflow scheduling based on evolutionary state controlled multi-objective particle swarm optimization algorithm is proposed.The scheduling algorithm proposed in this thesis can obtain the optimal scheduling scheme set closer to the real Pareto front and can jump out of the local optimal state while accelerating the algorithmâ€™s convergence speed.First,the ESCMOPSO-CWFS algorithm introduces the concept of interquartile range statistics and information entropy to design a quantitative assessment method for particle swarm status.This method enables dynamic monitoring of the evolutionary state of the particle swarm.Then,in order to solve the problem that the particle swarm may converge to the local optimal state in the process of evolution,a perturbation strategy is introduced to apply disturbance to the process of particle swarm evolution,so that the particle swarm can change the convergence state and explore in a scheduling scheme solution space,so the obtained scheduling scheme set is closer to the true Pareto optimal front.Then,the algorithm also designed the dynamic maintenance strategy and global optimal particle selection strategy of the archive,which not only promoted the particle swarm to accelerate to the potential optimal solution space,but also ensured the diversity of the final candidate scheduling scheme solution set.In the end,this thesis expands the simulation platform Workflow Sim,and uses the practical workflow application instances to simulate the proposed scheduling algorithm on the extended platform.The experiment compares the scheduling algorithm proposed in this thesis with PSO-CWFS and MOPSO-CWFS algorithms.The results show that the scheduling solution obtained by the algorithm has faster completion time and lower execution cost,and the candidate scheduling solutions are more evenly distributed,which gives user more options. |