Research On OpenStack Distributed Virtual Routing And Bandwidth Allocation Strategy

Cloud computing is a new delivery model and business model for computing resources,network resources and storage resources,and enhances services through dynamically scalable virtualized resources.As an open source cloud platform project with high activity,OpenStack receives wide attention from the business community and academia for its efficient community development and flexible deployment model.At present,most of IT vendors adopt the traditional network architecture to deployment OpenStack.The centralized network service makes network nodes vulnerable to network bottlenecks such as single point failure and excessive traffic load.Therefore,how to decrease the traffic of network becomes one of the key issues of the cloud platform.The instances are provided by the OpenStack cloud platform.They are used to the user as a deployment basic application service,which provide single application service.And a part is provided to the user as a tenant virtual machine,which will deploy multiple applications.As the demand of users increases,the number of the application service virtual machine and tenant virtual machines used will increase,the amount of stored data will become larger and larger,and the competition for cloud platform network resources will become larger and larger,too.In order to solve the problem of data storage maintenance,the cloud platform has adopted Ceph to implement a back-end unified block storage system,and the created instances are scheduled on each computing node to achieve load balancing.For the competition problem of network resources,static or dynamic bandwidth allocation can be used to solve the problem.However,the static allocation mode has low resource utilization,and the traditional dynamic bandwidth allocation mode cannot guarantee the application service QoS.Therefore,how to distribute the bandwidth of virtual machine on demand is also one of the problems that the cloud platform needs to solve.In view of the above problems,this paper first studies the distributed virtual routing technology,and applies it on the OpenStack cloud platform,which implements the high availability of the cluster network and solves the performance bottleneck of the network node.Based on this,the bandwidth dynamic allocation strategy of single application virtual machine and multiple application virtual machine is researched.Researching the traffic characteristics of the single application service virtual machine,furthermore,proposing tolerant delay to ensure the improvement of effective bandwidth calculation method,overcome the shortcomings of only packet loss rate guarantee.After that using the ARIMA model to predict the network load status of the application service virtual machine in the future,thereby realizing the dynamic bandwidth allocation of the application service virtual machines.The experiments verify that the strategy has the advantage of guaranteeing the application service QoS compared with the traditional strategy.And then a spike detection algorithm based on z-score is proposed to detect the traffic fluctuation degree of the multiple application virtual machine,which realizes network QoS guarantee for multiple applications of virtual machines through loose throughput guarantee.Finally,the performance of the OpenStack cloud platform system is tested.The experimental results show that the improved cloud platform network performance has been greatly improved.