Research On Network Traffic Measurement Technology In SDN Environment

Traffic is the main carrier of network information circulation,and the characteristics of traffic reflect the behavior of the network.With the surge of network traffic,network management and monitoring have put forward higher requirements for network traffic measurement.Network traffic measurement plays an important role in network management and monitoring,and is the cornerstone of network management and monitoring.The measurement information provided by network traffic measurement can ensure the normal progress of various network tasks,including load balancing,traffic engineering,and anomaly detection.In recent years,new network technologies such as SDN and P4 have brought flexible control planes and programmable data planes,providing new design ideas and experimental environments for network traffic measurement.Although the programmable data plane makes traffic measurement methods no longer restricted by traditional network hardware,how to quickly process large and complex network traffic in limited memory space to complete important network measurement tasks is still a research in the industry and academia.This paper first conducts a series of feature analyses and mathematical modeling on real network traffic,focusing on the distribution of elephant flow and mouse flow in the traffic,as well as the fitting verification of the heavy-tailed distribution of traffic,and the large The data packet arrival interval of the elephant flow is discussed,and the conclusion that the data packets of the elephant flow presents a concentrated distribution in time is drawn.Then,the advantages and disadvantages of the existing top-k elephant flow detection methods are analyzed,and a top-k elephant flow detection method that is more suitable for traffic characteristics is designed to address the shortcomings of the existing methods.Compared with the existing methods,this method not only distinguishes the mouse flow from the elephant flow through the size characteristics of the flow,but also preserves the local time characteristics of the flow in a limited space through a new data structure,so that it is more effective The record of the newly arrived flow is saved,so that the newly arrived flow can have enough time and space to grow into an elephant flow,and the mouse flow that cannot become an elephant flow can be discarded in limited memory space in time,to accurately retain Information on elephant flow.Finally,a large number of parameter experiments and comparative experiments are carried out on the real backbone network traffic data set,which proves that compared with the current optimal method,the top-k elephant flow detection method designed in this paper has higher accuracy.Lower errors and higher throughput.Based on the designed top-k elephant flow detection method,this paper develops a set of in-band network telemetry systems based on elephant flow detection.With the help of the elephant flow detection method locally on the device,the information of the elephant flow is effectively recorded,and the information of the elephant flow is inserted into the header of the service data packet,and the more critical telemetry information of the elephant flow is carried by the service flow,to guarantee Under the condition of low bandwidth consumption,it can effectively complete network-level measurement tasks such as network black hole detection,network loop detection,and abnormal jitter detection.