Research On Workflow Scheduling Algorithm Based On Failure Aware In Cloud Environment

In recent years,due to the rapid development of cloud computing,large-scale cloud computing data centers are widely established around the world.With the increase attention in cloud computing,its functionality and complexity are extensively studied.In order to solve the premise of how to meet the quality of service,as much as possible to reduce service costs,to achieve both users and service providers to maximize the benefits in cloud,the concept of"Cloud workflow" could be introduced.Through the combination of workflow technology and cloud computing,on the one hand,it can transfer the complex application requirements in accordance with the business logic to the abstract definition and decomposition of the task,resources re-order and flexible configuration,thereby improving the service quality;on the other hand,it can realize the task automatic scheduling,task monitoring and resource allocation optimization and management,so it can greatly improve the efficiency of implementation the task,can effectively improve the quality of cloud computing services,reduce the cost of the implementation of the task burden.In order to solve the scheduling problem of cloud workflow,different researchers have studied from different views:such as minimizing the completion time,minimizing the implementation cost and maximizing the task completion rate,etc.Although there are many scheduling methods of cloud workflows,none of the scheduling models are established for the failure of cloud computing resources,so as to effectively avoid and reduce the impact of failure events on workflow task scheduling results in cloud environment.In the cloud computing environment,resource failure is inevitable.Due to resource failure will directly bring the system performance degradation,the implementation of the program early termination or even data loss and other problems,eventually leading to more tasks can not be completed within the deadline,increased default rate,seriously affecting the reliability and stability of cloud computing,and greatly reducing the quality of service(QoS).At the same time,there are timing constraints and data dependencies between the various tasks of the workflow,therefore,in the implementation of the workflow process,once a resource node failure,not only lead to the task needs to re-implementation,it is possible that the entire workflow tasks need to re-implementation,so it can be seriously affect the efficiency of cloud computing,and waste a lot of computing resources.It is a development trend of domestic and foreign research on failure prediction mechanism in current cloud computing environment.Combined with the characteristics of cloud computing scheduling optimization,for the first time,this paper proposes a workflow scheduling model based on failure aware mechanism,and introduces the failure prediction mechanism and task re-scheduling strategy,to meet the cut-off period on the basis of the task to maximize the completion rate as the goal.In the task scheduling process,for each task of the workflow to generate sub-deadline.According to the resource failure prediction model,when the selected resource node fails before the deadline of the task,the task is migrated to another node which can successfully complete the task,so as to effectively avoid the resource failure bring to the task and allocate critical path tasks to the same high-performance virtual machine as much as possible during mission migration,it can reduce the overhead of data transfer between tasks,shorten the completion time and improve the completion rate of the task.Then,the function of each module in the failure-aware workflow scheduling model is described in detail.After that,the failure prediction mechanism,the workflow model and the resource model are defined,and the model is implemented.A workflow algorithm based on failure aware(BFGA)is proposed.The algorithm is based on the improved genetic algorithm,proposed a new triple ternary coding method,in the population initialization process,the use of random generation and use has been proved to be effective combination of algorithms to generate individuals in order to achieve the diversity of population of the goal.In this paper,we design the cross and mutation method which accord with the characteristics of workflow,and then introduce the adjustment operator to fine tune some of the results after the cross mutation of the individual,so as to avoid the local optimum and improve the convergence speed.Simulation experiments of the proposed models and algorithms are carried out through CloudSim cloud computing platform.Experiments are carried out by means of different types of workflow applications and changing of simulation environment parameters.By comparing with the GA algorithm,the validity of the algorithm is verified.It is proved that the BFGA algorithm compared with the general GA algorithm.,by using the triplet coding method,the initial population adopts a variety of methods to generate the individual,which enriches the diversity of the individual population and adds the adjustment operator to the population evolution process,has a better convergence rate.Secondly,the effect of BFGA algorithm and GA algorithm,First-fit algorithm,Pessimistic Best-fit algorithm and general algorithm without task on task scheduling is verified from the three aspects of failure prediction accuracy,workflow task number and failure node ratio.The experimental results show that the BFGA algorithm has higher operation completion rate and unreliable node utilization rate than other algorithms when the failure prediction accuracy is greater than 50%.When the accuracy of failure prediction is 75%and the number of workflow tasks is greater than 600,the task completion rate of five algorithms decreases,but the BFGA algorithm is slow and has a higher accuracy than the other four algorithms.The experimental results show that the BFGA algorithm can reduce the impact of resource failure on workflow scheduling,which successfully solves the problem of workflow scheduling based on failure aware.