Research On The Key Technology Of Dynamic Optimal Scheduling Of Multidimensional Resources In Cloud Computing Environment

Dynamic resource provision based on virtual technology is one of the core features of cloud computing system,which brings "Pay as you go" service mode to cloud computing and enables elastic and efficient resource management of cloud environment on the basis of resource pooling.Facing dynamic workload,the variety of task requirements,as well as heterogeneous,multi-dimensional cloud resources,how to further improve the efficiency of resource utilization,how to balance the multi-dimensional cloud resources,how to ensure reliable task execution of workflow in the cloud environment has been becoming the current cloud computing scheduling optimization research institute of several typical challenges,in this paper,starting from the dynamic characteristics of the cloud,summarize its impact on the cloud resource scheduling and puts forward the corresponding solutions.In this paper,the dynamic characteristics of cloud environment are summarized as the following:1.Non-stability,which refers to the unstable status of cloud resource utilization caused by dynamic task changes,specific performance in unable to use the mathematical model of stable characterization of cloud change law of the host load,and dynamic changes of load affects the effect of resource scheduling,to further improve the efficiency of resource utilization.2.Unbalance,it refers to the non-equilibrium of different kinds of cloud resources caused by the diversity and dynamics of mission requirement,the non-equilibrium characteristics embodied in a cloud environment are expressed in large difference between host all kinds of resource utilization level,which may appear certain resources exhausted other resources still have more rest but can't be used again.3.Unreliability,it refers to by the cloud environment caused by the extensive use of consumer server host fault probability,system reliability,the manifestation of node failure become "normal state" of the system,with the high service quality or sensitive to fault,the workflow task will be affected by the larger,lower cloud system performance.Based on the above analysis,this paper conducts an in-depth study on the dynamic characteristics of cloud environment and the challenges brought by the research on scheduling optimization.The main work and innovation points of this paper include:(1)Research on the Decompose-Compose with Error Compensation(DCEC)workload prediction algorithm in cloud resource schedulingAgainst the unstable characteristics of dynamic changes in resource utilization in cloud environment,low prediction accuracy and bad prediction algorithm instantaneity in cloud workload prediction research,a dynamic workload prediction online algorithm which is named as Decompose-Compose with Error Compensation(DCEC)Workload Prediction Algorithm is proposed in this paperr.In the algorithm host workload is decomposed to several independent task workload,an improved Kalman filter is used to predict the task workload,and an error compensation mechanism based Markov chain model is added to the algorithm for correcting the final prediction results.The proposed algorithm has advantage of low computation cost and high prediction accuracy.Several experiments compare to classic workload prediction algorithms show that the proposed algorithm is more efficiently and accurate.(2)Research on multi-dimensional resource scheduling based on 2D-BPSO virtual machine and container combinedIt is aimed at the unbalanced utilization of various resources in the cloud environmnent,with the increasing growth of applications deployed to the cloud computing platform,cloud providers have made a higher request for cloud resource utilization eff iciency,but there is bottlenecks in cloud service provisioning base on VM.Against these problems,we proposed a VM-Container hybrid hierarchical resource scheduling mechanism which classified tasks into different levels according to the comprehensive of virtualization type and scheduling state.By using the characters of lightweight and easy to migrate,we formulate the initial scheduling strategy for tasks in different levels and design the combination optimization scheme when there is a migration of allocated task.We optimized the scheduling problem with PSO,and expand the traditional 1D-PSO to 2D-PSO and give a general formulation of nD-PSO for further,we proposed a binary-operator-based iterative update strategy to form the binary-operator-based 2D-BPSO to solve the optimization problem.We designed and implemented simulation program,the simulation results verify the feasibility of the proposed algorithm which can enhance the resource utilization efficiency in this simulation.(3)Research on reliable workflow task schedulingAgainst the non-reliability characteristics of node failure becoming normal in the cloud environment,this paper selects workflow task scheduling problems with high requirements for system reliability to be studied.Traditional research often takes fixed structure workflow as the research object,and set the task completion time or cost as the optimization goal,using static scheduling method as the core algorithm of workflow scheduling problem is studied.But in a cloud environment,species diversity,flow processing workflow tasks become the main object of scheduling,meanwhile the virtual machine node failure or server node failure will have important influence on a dependency between workflow tasks,this paper put forward a reliability oriented workflow task scheduling mechanism,using the directed acyclic graph(DAG)for workflow modeling,workflow task fault-tolerant scheduling with prime-backup copy technology,in this paper,on the reliable cloud environment,improve the classic fault-tolerant scheduling algorithm FT-HEFT,HEFT algorihm is derived theoretically to realize fault-tolerant scheduling needs to satisfy the constraint conditions,based on the typical workflow in reality,to simulate the flow processing application scenario simulation experiment,the experimental results verify the feasibility of the proposed algorithm,on the premise of guarantee the reliable task execution,optimize the workflow task completion time and cost,and in contrast with other scheduling algorithms to verify the algorithm optimization results,It has guiding significance for fault tolerant scheduling of workflow tasks in cloud environment.To sum up,this paper takes the instability,non-equilibrium and non-reliability of cloud environment as the research entry point and proposes corresponding solutions to the problems brought by each point.According to the different dynamic characteristics of different cloud environments,the research contents of this paper can be applied independently or in cooperation with each other,which is the development direction of the follow-up research.Therefore,the research of this paper can play an active role in the research field of cloud environment resource scheduling.