Research On Virtual Machine Scheduling Algorithm And Implementation Of Evaluation Tool

In recent years,with the popularity of cloud computing services,the scale of data center is growing.Meanwhile,large energy costs and low utilization of server resources has become an increasingly serious problem.The virtual machine scheduling is the key to solve this problem.Most of the proposed virtual machine scheduling algorithms are optimized for single or dual targets.In practice,virtual machine scheduling needs to consider multiple objectives,such as reducing energy consumption,ensuring cloud service quality,load balancing,etc.In addition,the verification and evaluation of virtual machine scheduling algorithms are mostly carried out by CloudSim simulation software.However,due to the shortcomings and limitations of CloudSim,the simulation and evaluation of virtual machine scheduling are more costly and difficult to use.In this thesis,a virtual machine scheduling problem is analyzed and studied deeply.Firstly,this thesis analyzes the virtual machine scheduling model in real cloud environment,and summarizes the general process of virtual machine scheduling.It is divided into three steps:the determination of migration time,the selection of virtual machine to be migrated and the placement of virtual machine.Then,a virtual machine selection strategy based on TOPSIS multi-objective decision-making is proposed and designed for the analysis and research of the selection problem of the virtual machine to be migrated.This strategy can balance the multiple targets in the process of virtual machine selection comprehensively.In the end,based on the limitations and shortcomings of CloudSim,CloudEval is designed and implemented,which focuses on the simulation and evaluation of virtual machine scheduling algorithms.The virtual machine scheduling algorithm proposed in this thesis utilizes the TOPSIS multi-objective decision-making method to balance the energy consumption,load balance and migration time.It can better manage the resources in the data center and realize the maximization of the benefit and performance.At the same time,CloudEval adopts a hierarchical architecture,which is based on the discrete event simulation engine.Simulation of each component in the data center is carried out.The simulation software is tested by implementing the proposed scheduling algorithm.Compared with CloudSim,CloudEval designed in this thesis is more in line with the characteristics of the data center.It supports the simulation of virtual machine live migration,optimizes the discrete event processing and is more scalable.So the researchers do not need understanding of the framework itself too much,and only need to do simple configuration to complete a simulation process.