Research On Network Virtualization Platform And Application Development

Academia proposed network virtualization technology to solve network rigidness problem. This technology in accordance with the idea of the server virtualization reconstructs the whole network, to create multiple virtual networks, while each virtual network can be deployed independently and have its own management system. Network virtualization as a support of existing cloud computing, whether in business or academic fields, have become increasingly important. Iaa S、PaaS、SaaS and other relying on network virtualization public services make cloud platforms attract more and more people to deploy services on them.The core issue of network virtualization is to complete virtual network mapping problem. Therefore, the first issue of this paper is make comparative studies of the existing network virtualization platforms(FlowVisor、OpenVirtex、CoVisor). In this study we will focus on the next four indicators: platform in network access point to identify the traffic for different tenants, and during traffic transmission to identify the traffic belongs to which tenant; and ensure the virtual node have forwarding table isolation; and network topology virtualization and isolation. After that, we will adopt experiment to test these platforms and evaluate the advantages and disadvantages of these platforms in these four technology.However, in actual production, only complete virtual network mapping can not meet the virtualization demands. Because in multi-tenants virtualized network environment, there must be competition for resources among tenants. Network virtualization platform also need to address the guarantee of the network resources. The key point of network resource guarantee is to allocate bandwidth resource. In order to solve this problem, this paper proposes a fair bandwidth guarantee algorithm for multi-tenants. The solution take bandwidth guarantee, maximize resource utilization and fairness these three aspects in account and achieve a real-time fairness bandwidth guarantee with efficient traffic scheduling scheme for multi-tenants. At the same time, we selected two existing multitenant bandwidth allocation scheme, FairCloud and ElasticSwitch, as a comparison of the algorithm. The result of comparison show that this algorithm can ensure the maximum use of network resources, while provide a fair allocation of bandwidth resources for tenants.Finally, we have constructed multi-tenant fairness bandwidth guarantee prototype system. Because of the lack of some features in network virtualization platforms, we designed a network traffic statistic module with adaptive traffic collection algorithm and take advantage of the Linux TC module to implements the QoS module in prototype. After that we make our bandwidth algorithm as an application embedded in system, and deploy the prototype in a real network environment to make a measurement. According to the result, the prototype system have viability in a real environment.