The Research On Traffic Measurement And Analysis For IP Network

Network traffic is one of the important metrics for describing network behaviors. It plays an important role in network design, service arrangement, network protocol and equipment design and traffic project implementation. This paper analyzes the current research and correlative applications of the IP network measurement technology, completes a distributed traffic monitoring system, and do a great deal of experiments on it. This paper constructs a traffic model for traffic prediction by analyzing experiments information.The main achievements during the period of Master degree are described as the following:Application requirement of IP network monitor and research of its current situation have been analyzed. This paper describs the requirement of IP network monitor technology research and deployment of monitor system from the IP network protocol point of view. It points out that the behaviors of network based on IP network monitoring are important for network protocol design, network capacity plan and design, network maintenance and providing network value-added services.Research and development on distributed IP network traffic monitor and analysis system. This paper tells a detailed distributed system design scheme. The architecture of this system is distributed collection and centralized storage mechanism. It can detect network traffic in real-time, and provide network service threshold alert, design, optimization, help for decision-making through the powerful expert analysis and event correlation. This makes network efficient, steady, secure for IP network system and ensures the key tasks completed successfully. The system is composed of monitor probes and a control analysis center. This paper focus more attention on design and implementation of monitor probes, the construction of probes, software layer, monitor task, deployment, task schedule, protocol analysis procedure, metric storage and sending etc.Built a traffic model for traffic prediction. By analyzing the other related network traffic models, a conclusion has been drawn that network traffic has Self-Similar characteristic but it make no use in traffic prediction. This paper achieved a traffic model based on the traffic data captured on real network. This model divides traffic into two parts: time related part and time unrelated one. The model's validation has been proved on the experiment data come from WAN and LAN. Based on the traffic model, a network traffic prediction algorithm has been proposed.