Research On Network Measurement And Performance Optimization In Software Defined Network

The rapid development of 5G technology has inspired a large variety of new applica-tions that can provide ubiquitous services to make the lives of existing industrial systems and people more intelligent.However,a large number of applications rely on a more in-telligent operator network,and the two main services that the operator networks provide are the Internet of Things(IoT)and broadband.However,under the traditional network architecture,the management of operator networks is very difficult,which breeds a new network technology,Software Defined Network(SDN).The centralized control of SDN greatly reduces the difficulty of network management,improves the support for cloud computing,virtualization and other emerging technologies,thus enhancing the support ability and service ability for the business carried by the network.In this paper,the soft-ware defined network is applied to the operator network.Firstly,We design and implement the software defined network measurement system based on sFlow,which can monitor the network traffic from different granularity in the operator network in real time,and then optimize the performance of the software defined network to improve the service ability of the operator network.This paper is mainly divided into three parts.1.Network measurement is not only an important way to deeply understand the net-work,but also a prerequisite for network optimization.We design and implement a net-work measurement system based on sFlow.The system is mainly used to monitor the network traffic data in three levels: port,host and application.The system is mainly composed of three modules: acquisition module,storage module and display module.The function,design and implementation of each module are introduced in detail.Finally,we evaluate the feasibility of software defined network technology and the correctness of measurement results in virtual network and real network.2.For the applications of IoT,the operator networks face the challenge of accesses of massive devices with the rapid development of 5G technology.However,a single con-troller in SDN has a limited request-processing capability,so a distributed control plane with multiple controllers has been used to achieve scalability and reliability for support-ing the IoT applications.In this paper,we investigate the domain partitioning problem in large-scale network with multiple controllers.Firstly,we verify the limitation of single controller's processing ability through experiments,and use hypergraph to analyze the non-uniform data transmission over multiple flow paths.Then we propose a Partitioning Algorithm for Software-defined IoT Network(PASIN)to achieve the minimum total load of requests with the constraint of total switches-to-controller delays in each domain.Our extensional simulations verify the effectiveness of our proposed scheme.3.For the applications of broadband,operators need to ensure the Quality of Service(QoS)of broadband services,and bandwidth is an important indicator of QoS.We inves-tigate the dynamic bandwidth allocation problem caused by the dynamic connection of hosts in the operator network.This paper analyzes the cascading problem for the dynamic rate in tree network topology.By considering the tradeoff between the performance and the cost of rate allocation,we formulate the rate allocation as an optimization problem,and propose a Rate Allocation algorithm with Limited Range in SDN(RALR).We adopt theory of Lyapunov drift to discuss the allocation range.Our simulation results verify the performance and efficiency of the proposed algorithm.