Research On Task Scheduling Method Based On Mixed Tasks In Cloud Computing

As one of the core technologies of cloud computing data center,task scheduling has always been the focus of research in academia and industry.With the continuous growth of global demand for cloud computing,the construction and deployment of cloud computing data centers continue to expand,and the problem of resource utilization of cloud computing data centers becomes increasingly prominent,which has become the focus of cloud providers.Development,application and the application of diversity resource requirements also showed a trend of diversified development,batch processing application in a large number of dependent tasks with different requirements and independent tasks coexistence posed a challenge to improve the utilization rate of resources,how to reasonable scheduling batch task to improve the utilization rate of resources,saving energy costs is of great significance.Based on the above background,this paper takes improving resource utilization as the optimization goal,and focuses on the research of batch task scheduling method.The specific work mainly includes the following three aspects:1.A method of dependent task scheduling based on table scheduling strategy is designed.In order to shorten the scheduling length of dependent tasks and take into account the heterogeneity of resources in cloud computing environment,the priority calculation method of table scheduling is improved.Then,based on the shortest completion time first algorithm,resources are allocated to the selected computing nodes of dependent tasks.In order to integrate the fragmented resources generated in the process of resource allocation,a task replication strategy with task lock as the core was proposed.A lock was added to each task to monitor the task allocation state,and the centralized execution of each precursor task was postponed without extending the length of job scheduling.Experiments show that this method can shorten the task scheduling length effectively.2.An independent task scheduling method based on null policy is designed.The method optimizes the idle resources generated in the task allocation process through idle resource monitoring,independent task sorting and idle resource allocation,so as to improve the resource utilization rate.Firstly,based on the strategy model,the spatial and temporal information of idle resources is dynamically monitored under three conditions: normal task allocation,task replication and idle resource reallocation.In order to reduce the time complexity of independent task assignment,the maximal matching principle and the shortest task priority principle were adopted to dynamically sort the independent tasks before task assignment,based on the three situations of reasonable task assignment.Then,in order to avoid the problem that the scheduler of the independent job is too long in the extreme case,the balance factor is used to calculate the priority of the independent task and select the task to be allocated to the idle resource according to this,so as to complete the allocation of the idle resource.Experimental results show that this method can effectively improve the overall utilization rate of resources.3.A hybrid task scheduling method,which can schedule both dependent tasks and independent tasks simultaneously,is proposed.In order to solve the interaction between the two kinds of tasks and reduce the problem of system performance,based on the difficult before they are easy,first distribution depend on the task after the thought of independent tasks,by using the proposed dependent tasks and independent task scheduling algorithm,and pieces by integrating resources and make full use of idle resources,to enhance the utilization ratio of resources,at the same time balance factor is introduced to prioritize,The overall task scheduling length is optimized.Experimental analysis on Cloud Simpy simulation platform,which supports static scheduling and dynamic idle resource monitoring,proves that the proposed method can improve resource utilization and shorten the average job scheduling length at the same time.