Research On Task Scheduling Strategy Of Heterogeneous Distributed System Based On Multi-DAG Model

In recent years,with the rapid development of communication and computer technology,the big data center platform built by large-scale distributed computing environment has been widely used in electronic commerce,industrial control,scientific computing and other fields.The excellent task scheduling strategy directly affects the computational efficiency and quality of the system.Recently,research on scheduling problems of single DAG task(directed acyclic graphs)on multiple computing resources has basically matured.With the development of computer technology and the continuous improvement of user requirements,the problem of multiple DAGs scheduling in heterogeneous computing environment has attracted wide attention.The problem of resources has gradually received wide attention,and related issues such as scheduling fairness and reliability have also become research hotspots.Therefore,this paper studies the problems related to multi-DAG task scheduling in heterogeneous distributed computing environment.The main work and innovations are as follows:(1)Research on the fairness of multi-DAG task scheduling.In this paper,a heuristic fair scheduling algorithm,IFairness algorithm is proposed.Based on the foundation of the fair scheduling algorithm,instead of residual Makespan,a new evaluation index DAG Completion Degree is chosen as the basis of DAG selection when selecting the DAG to be scheduled,furthermore,in the calculation of slowdown for each DAG,the principle of "backward looking" is adopted to solve the problem that some DAG can not be scheduled in the early stage.Simulation results show that IFairness algorithm effectively improves the fairness and resource utilization of the scheduling algorithm.(2)Research on reliability issues in task scheduling.Since the computing power and the probability of failure of the computing nodes in the heterogeneous distributed system are not necessarily the same,the task and the processor are firstly modeled for reliability.Considering the task fairness,task constraint relationship and computational node failure probability,this paper proposes a multi-DAG task scheduling strategy Reliable Fairness Algorithm which drives the task fairness with reliability target and makes full use of resources.Ensure the efficiency and reliability of the system.Finally,the experimental system is designed to simulate the performance of the algorithm.(3)The multi-objective collaborative optimization problem of reliability,scheduling efficiency and fairness of multi-DAG task scheduling is studied.According to the previous simulation,the results obtained by the traditional heuristic scheduling algorithm are not optimal solutions in the case of known tasks and system models.Therefore,this chapter proposes a multi-objective optimization based on cuckoo search algorithm(MOCSA).Firstly,the content of the cuckoo algorithm is introduced.Then the proposed algorithm is designed and the core code of the algorithm is given.Finally,the experimental system is designed to simulate the performance of the algorithm.