Research On The Reliability Scheduling Mechanism Of Cloud Workflow Based On Dynamic Backup

As the third revolution of the Internet industry,cloud computing is a big leap in the information age and will play a pivotal role in the future.More and more researchers are studying on cloud computing to promote the rapid development of cloud computing technology.Among the research on performance indicators,the reliability research of task execution is still lacking.However,with the continuous development of cloud computing,the system reliability has also been an important factor,which is related to performance indicators such as execution time and cost.Currently,most of researches use static replication and fault-tolerant methods to statically improve the system reliability.But the static method cannot fully utilize the resources of the cloud computing platform,which will inevitably cause the waste of resources.Therefore,dynamically improve the reliability is of great significance for task execution in cloud computing platforms.Combines dynamic replication method,this paper study on the completion time and system reliability indicators that affect workflow scheduling in the cloud environment.Among them,time is a constraint,and system reliability is an optimization objective.The main work and research contents of this article are as follows:(1)This paper analyzes the characteristics of the cloud environment and scientific workflow,model the current workflow and establish a heterogeneous model in the cloud environment,then give the time and system reliability calculation model.(2)A DCDR(Deployment Cloud computing Dynamic Replication)algorithm is proposed to optimize the reliability of the scientific workflow task execution system with deadlines as constraints in a cloud environment.The algorithm includes the design of execution order of workflow tasks,matching the workflow tasks according to the sequencing order,the resources matching of the cloud environment,and improving system reliability by combining the idea of dynamic replication.This paper gives simulation experiments by a large number of scientific workflows based on real applications data sets.The experimental results show that the DCDR algorithm can improve the reliability of the system effectively when meeting the deadline constraints.(3)This paper proposed a time-reliability optimal dynamic scheduling algorithm MRC(Makespan Reliability optimization for Cloud)for scientific workflow tasks execution in the cloud environment.Based on the DCDR algorithm,this algorithm introduces the concept of fuzzy advantage to prioritize tasks,and uses Pareto technology to demonstrate the multiobjective optimization problem of task scheduling in the cloud environment.Experimental results show that the MRC algorithm has a better Pareto frontier than SPEA2 * algorithms on a variety of scientific workflow data sets.