Research On Key Issues Of Applying Software-Defined Networking To Multi-Tenants Datacenter

As a consequences of the explosive cloud applications and users,Datacen-ter Networking(DCN)has become an essential component of cloud ecosystem which supports tasks requiring tremendous computing.However,traditional DCN architecture is challenged in several aspects,such as performance guar-antee,security enforcement,and resource and energy management.Especially,under multi-tenant circumstances,a flexible,secure and effective network ser-vice is an urgent demand.Software Defined Networking(SDN),as an emerging network architec-ture,with its features of centralized management and network programability,not only improves performance of network,but also provides novel network-ing functions and improves network resource utilization.Due to its flexible architecture which can meet the requirements of DCN for complex cloud ap-plications,applying SDN to DCN has drawn great attention from academia and industry.For the software-defined data center networking,this paper studies three aspects:data plane connectivity,control plane east-west data management,and network transmission scheduling across DCNs.The main contributions of this paper are listed below:(1)Aiming at the network connectivity problem,this paper proposes an ac-tive network connectivity fix strategy.With the migration of virtual machine,the state of the data plane changes.The network configuration pushed by the controller before the migration is not functional after the migration,which is-sues the problem of the network connectivity.Firstly,this paper proposes adopt the network update model to the scene of virtual machine migration to formalize the network state and configuration relationship,and gives formal definitions to virtual machine migration and network connectivity.On the basis stated above,this paper proposes Active Network Connectivity Fix Strategy.It calculates a set of network configuration rules that need to be updated by comparing the two network states before and after the migration,and actively updates the network as soon as possible when the migration occurs to recover connectivity of the data plane.The simulation results show that the active network connectivity fix strategy can recover the network connectivity at a faster and stable speed,meanwhile occupying less data plane resources.(2)Aiming at the control plane east-west data storage and synchronization problem,this paper designs and implements a blockchain-based east-west data management platform.To enable data exchange across controllers,the develop-ers of SDN controllers design and implement incompatible east-west data syn-chronization mechanism and interfaces.However,the responsibilities of data storage and synchronization are still taken by the controllers,which may de-crease the stability of DCN when one of the controllers is not functional.Firstly,this paper proposes a business model for data storage which enables users ex-press their demands on the storage performance.On the basis stated above,this paper designs and implements a blockchain-based east-west data management platform.It can make full use of the blockchain technology to transfer data synchronization,update and verification tasks from the controller to the plat-form,thus reducing the load on the controller and improving the reliability of the control plane.Also,this paper expresses an actual use case which includes two functions,virtual machine network address management and connection-less remote interface invocation,to demonstrate the functions and features of the proposed platform.The simulation result shows that the data management platform not only flexibly meets different requirements of data storage,but also reduces the load of controllers and provides secure,reliable data service.(3)Aiming at network transmission scheduling across DCNs problem,this paper proposes a completion time-based bandwidth scheduling strategy.Under the condition of limited network bandwidth,ensuring the network transmis-sions can be completed before the deadline will directly affect the performance of various services running by the tenants across DCNs.Firstly,this paper pro-poses a completion time-based transmission model which enables tenants de-fine their transmission tasks and the impacts of different completion time.On the basis stated above,this paper proposes a completion time-based bandwidth scheduling strategy.It allocates bandwidth by predicting the completion time of different transmission tasks to find a maximum utility within all possible so-lutions.Additionally,this paper introduces the fairness coefficient to adjust the fairness and interoperation of the allocated result.We also design and imple-ment the Inter-DC WAN simulator which is able to simulate arbitrary network topology and dynamically configure the status of links.The simulation result shows that the completion time-based bandwidth strategy can effectively im-prove the utilization of the network bandwidth and the total utility obtained by the system.