The Key Technology Research Of Resource Scheduling Based On OpenStack

Recently,cloud computing has been integrated into different aspects of modern computer systems.It is relevant to the emerging technologies,including 5G,Internet of Things,mobile computing,and big data analysis.As being an important supportive infrastructure,cloud computing has also become one of the key technologies in the future IT industrial business.However,several challenges appear with the development of cloud computing technology and the increase of the scale of the underlying infrastructure.Specifically,we focus on how to effectively schedule resources among computing nodes and improve the resource utilization at the same time.Such ability lay the foundation of guaranteeing the resource demanding requests from each user,and maintaining load balance on computing nodes,thus becoming one of the most important issues in the cloud computing research community.In view of the above challenge,this paper study the resource scheduling mechanism based on the popular OpenStack cloud platform.The major works are as follows:1.We analyze the resource scheduling schemes of OpenStack,and point out the drawbacks existed in resource mapping and virtual machine migration.To overcome these drawbacks,this paper propose OTRSM,a novel resource scheduling model based on OpenStack.OTRSM divides the scheduling process into two stage: virtual machine mapping,and online migration.Based on the two stage design,the number of physical machine in use is reduced,and the available resources are fully utilized,which turn out to effectively reduce the waste of resource on physical machine.2.On the mapping stage of Virtual Machine,we propose an improved quantum-behaved particle swarm optimization algorithm,AQPSO-M.According to the relationship between particle's current position and the optimization position,AQPSO-M adaptively adjusts the contraction-expansion coefficient and the particle weight.Simulation results show that our algorithm can improve the efficiency of resources allocating and reduce the system energy consumption with all users' requirement fulfilled.3.On the live migration stage of Virtual Machine,we propose MinLPF-LM,an algorithm to evaluate the load pressure of a node based on the load size of each physical node and the evenness of resource utilization on each dimension.We then leverage these information to maintain the load balance of among physical nodes and resource utilization,and perform a load-minimum virtual machine migration.The experiment results show that MinLPF-LM can accurately reflect the evenness of resource utilization on physical nodes,and effectively improve the resources utilization.In summary,our work optimizes the existing resource scheduling policy on OpenStack from three aspects,including the load pressure on nodes,the evenness of resource utilization on physical nodes,and energy consumption,that can make the resource scheduling on OpenStack more reasonable and effective,and has the good practical significance.