Research On Task Scheduling Optimization Oriented To Cross Microservice Chains

The application is divided into a set of loosely coupled fine-grained services by the mi-croservice architecture.As the services cooperate with each other,multiple intersecting mi-croservice chains are formed.A reasonable allocation method for resources not only solves the problem of resource competition caused by service chain crossing,but also improves resource utilization and reduces task response times in the process of task scheduling.However,the existing studies tend to ignore or simplify the conflicts that occurs when cross-chain access microservices,which leads to poor scheduling effects of the system.Aiming at these problems mentioned above,this paper conducts optimization research from two aspects of system modeling and scheduling algorithm.The specific work is as follows:(1)In order to accurately predict the resource consumption and task execution time overhead in the microservice architecture,a system resource model and task response time model based on the microservice chains are constructed;the relationship matrix is used to represent the invocation relationship of the microservice chains in the system.Considering the utilization of system resources and the global response time of processing requests,a multi-objective optimization model based on microservice chains task scheduling is constructed,which Provides accurate model support for the subsequent scheduling algorithms and the system to achieve efficient and accurate cross-chain scheduling management.(2)In order to solve the problem of resource competition and long response time in cross-chain scheduling of traditional scheduling strategies,a Chain-oriented Task Scheduling Algorithm(COTSA)for microservice based on the above multi-optimization model is proposed.This algorithm combines the parallel computing advantages of the ant colony algorithm and the strong local search ability of the simulated annealing algorithm.First,the tree structure is used to generate the solution space of the microservice executable sequence.Secondly,the quality evaluation function is used to measure the superiority of the feasible solutions to ensure the effectiveness of updating the ant colony pheromone matrix.Finally,according to the multi-objective heuristic information in the iterative process,the optimal path selection probability is improved.At the end of each iteration,a neighborhood solution space is generated according to the current feasible solution and the strong local search ability of simulated annealing algorithm is used to generate a new feasible solution in this space.If it is better than the current optimal solution,the current solution will be updated,otherwise,the new solution with a certain probability be accepted to avoid the ant colony algorithm falling into the local optimum.This thesis expounds on the influence of the system model and key parameters of COTSA on the optimization goal and algorithm performance through experiments,and the experimental results are analyzed.COTSA is also compared with FCFS(First Come First Service)and ACO(Ant Colony Optimization)algorithm.In terms of the resource utilization rate,COTSA is better than FCFS and ACO;In terms of the global response time of the task,COTSA is lower than ACO algorithm,when the number of requests is higher than 300,COTSA is superior to FCFS.The experimental results show that COTSA proposed in this paper has achieved superior results in solving the problem of multiple service chains competing for physical resources.