Research On Flow Management Strategy In Software-Defined Networking

The main task of network traffic management is to measure and analyze the information of network traffic,and design a reasonable routing scheme based on the information to optimize the network traffic distribution,which can improve the transmission efficiency of the traffic and the utilization efficiency of network resources.Switching devices forward traffic according to a predefined network protocol,making traditional networks inflexible and non-scalable.Software Defined Networking(SDN)emerged as new network architecture,having many advantages,such as isolation of control and forwarding,global centralized control,and programmability of network behavior.These distinctive characteristics of SDN greatly enhance the network's ability to manage traffic,so this thesis focuses on the network traffic management method based on the SDN.Network traffic information is an important input of routing optimization,including the information of flows and network status.The network status contains the round trip delay of the link,while the information of flows contains the size of the flows.Therefore,this thesis studies the traffic measurement and link round-trip delay measurement methods first,then we optimize the routing scheme according to the information to improve the utilization rate of network resources and achieve load balance.Round-trip link delay is an important indicator for network performance optimization and troubleshooting.This thesis proposes an efficient measurement scheme,which infers round-trip link delays from end-to-end delays of some measurement paths.The measurement paths are constructed by leveraging the flexible routing control capability of SDN.In order to reduce the measurement cost and meet the monitor capacity measurement constraints,the second chapter of this thesis studies the Monitor Placement and Link Assignment(MPLA)problem,and propose an efficient heuristic algorithm,MPLA Algorithm based on Biding Strategy(MPLAA-BS),to solve the problem.The size of network flows refers to the amount of data transmitted on the network,and is an important input for routing optimization.In order to obtain accurate network flow size information,this thesis proposes a network-level fine-grained flow measurement method based on traffic statistics.The method estimates the elephant flows and the mouse flows in the network by the statistical characteristics of the flow size obeying the normal distribution and then measures the estimated large flows.The experimental result shows that the accuracy of the proposed method is 84.0%.After measuring the convection,the size of all flows can be obtained,which can provide important information for route optimization.The main purpose of traffic management is to adjust the distribution of network traffic according to the network information and design a reasonable route optimization scheme to improve the efficiency of traffic transmission and the utilization of network resources.Therefore,route optimization is a very important aspect of traffic management.On the one hand,in an SDN switch,flows need rules for forwarding,and flow rules are generally stored in the TCAM,but the TCAM resources are limited;on the other hand,upgrading an existing IP network to an SDN network cannot be implemented immediately.Therefore,this thesis studies the routing optimization problem of hybrid SDN networks with TCAM capacity constraints.This problem can be expressed as a Mixed Integer Linear Programming(MILP)problem.This thesis proposes a variety of approximation algorithms that are solved in two stages.These algorithms can obtain good routing optimization performance by selecting the traffic path using a small number of TCAM entries in each SDN switch.