Research On OpenStack Platform Based Load Balance Technology

The rapid development of cloud computing has overturned the traditional IT architecture,which brings a new direction to the utilization of traditional physical resources.Cloud computing uses virtualization technology to virtualize physical resources into virtual resource pools,and allocates resources to users in a dynamic and scalable manner according to user needs.As the scale of cloud platforms continues to increase,simply expanding the scale of cloud data centers cannot solve the problems of low utilization,high energy consumption and unbalanced system load of data centers.In view of the above problems,based on the open source cloud computing platform OpenStack and combined with the existed cloud computing resource scheduling algorithm,this thesis conducted an in-depth study on the lack of dynamic resource scheduling in OpenStack.The main work of this thesis are as follows:Firstly,the concept of cloud computing and related technologies are studied,and the resource scheduling algorithm in the existing cloud environment is deeply analyzed.OpenStack platform and its related components are introduced,then the defects of the scheduling algorithm in the Nova module and the Swift module are pointed out.Secondly,aiming at the defects of the OpenStack platform scheduling algorithm and combined with the existed research,a grey Markov-based dynamic resource scheduling algorithm is proposed.This algorithm uses the improved gray markov model to predict the load information on the running nodes,and dynamically schedules the virtual machine by combining the future trend of the node load information with the dynamic scheduling mechanism of the virtual machine and the threshold of the upper and lower limits of the load information set in advance.The simulation results show that the grey Markov-based dynamic resource scheduling algorithm is more effective than OpenStack's default scheduling strategy and other algorithms,which improves resource utilization,reduces system energy consumption and achieves system load balance.Finally,in the cloud storage environment,aiming at the problem that the reading and writing speed of the system drops sharply and the load of the system is unbalanced due to the increase in the number of tasks,this thesis proposes the chaotic particle swarm optimization algorithm based on Swift.By improving inertia weight and learning factor in particle swarm optimization algorithm,the algorithm makes them change adaptively,enhances the searching ability of the algorithm and improves the searching precision.By introducing chaotic optimization method and precocious judgment mechanism,the phenomenon that particle swarm optimization is prone to fall into local optimization in the iterative process is solved.Simulation results show that the chaotic particle swarm optimization algorithm based on Swift can effectively improve the rate of reading and writing and achieve the system load balance in the case of high load.