| In cloud computing environments,applications are often organized and managed in the form of workflows.By scheduling workflows,tasks in the application can be assigned to the most suitable cloud resources to meet users’ needs for service quality.With the development of microservice technology,microservice architecture has become an important trend in the field of cloud computing today,and more and more enterprises are adopting microservice architecture to build applications.Therefore,microservice workflow scheduling has become an important research issue.However,most existing workflow scheduling algorithms are focused on traditional workflow scheduling problems.Compared to traditional workflows,the number of tasks in microservice workflows is larger,and the task execution environment is also different.When using traditional workflow scheduling algorithms to schedule microservice workflows,there are problems with low quality scheduling schemes and long generation time of scheduling schemes.In response to the above issues,this thesis studies two microservice workflow scheduling problems with different optimization objectives.The specific research points are as follows:(1)A list heuristic microservice workflow scheduling algorithm is proposed for the cost optimization microservice workflow scheduling problem under deadline constraints.The algorithm is divided into two stages: task priority sorting stage and resource allocation stage.In the task priority sorting stage,the algorithm first calculates the relaxation time and assigns sub deadlines for each task in the workflow,and then sorts the tasks based on the upward sorting value to form a scheduling list.In the resource allocation stage,the algorithm first searches for a set of available resources that meet the requirements for the tasks to be scheduled.Then,it combines the advantages of two existing resource selection strategies and proposes a dynamic resource selection strategy.Based on this strategy,the optimal virtual machine instance is selected for the task.The simulation experimental results show that the algorithm proposed in this thesis can achieve lower costs under different deadlines or different task quantities of workflows,while meeting deadline constraints.(2)A microservice workflow scheduling algorithm based on a joint hierarchical strategy and genetic algorithm is proposed for the time and cost dual objective optimization of microservice workflow scheduling problem.The algorithm is divided into two stages: task layering stage and task scheduling stage.In the task hierarchy stage,the tasks in the microservices workflow are divided into several task sets based on the hierarchy in the directed acyclic graph where the tasks are located.Layered operation ensures that there are no prior constraints between tasks in each layer,thereby maximizing the parallelism of tasks;Layered operations also reduce the search space of the algorithm during the task allocation stage,reducing the difficulty of algorithm optimization.In the task allocation stage,the Metropolis criterion from simulated annealing algorithm is introduced into genetic algorithm to enhance its local search ability,and an improved genetic algorithm is used to find the optimal scheduling solution for the task set.The simulation experimental results show that the algorithm proposed in this thesis can achieve a better scheduling scheme in a shorter time. |
