Design And Implementation Of Flow-Based IP QoS Traffic Analysis System

With the rapid development of the networks technologies, the increase of network services and the merging of networks, users have much more rigorous requirement for the quality of service of networks. The traditional best effort IP network cannot meet the requirement of various services. Then IP QoS technology became popular and many kinds of QoS technologies came out, to meet the new services development and QoS requirement. CNGI-QoS Management System combines the technologies of services management, bandwidth resources control, QoS mechanism on network devices and network operation, and provides a total QoS solution for next generation networks management.To provide better QoS for our users, the network administrators need network measuring information as supports for their daily network maintenance, and the Flow-Based IP QoS Traffic Analysis System is designed for this purpose specially. As an assistant system for CNGI QoS Management System, this system gives supports such as QoS problem detection, charging information support, services analysis and so on. For this system, the main technology used here is flow-based passive measuring, and with active measuring as assistant.This paper firstly introduces IP QoS technology and 3 kinds of popular QoS models, and the network traffic measuring technologies including relative standard and the application of traffic measuring. As the key technology, the detailed information of Flow Technology is discussed separately. After the architecture and key technologies of CNGI QoS Management System briefly introduced at the beginning, The Flow Based IP QoS Analysis System is brought forward as an improvement solution for CNGI QoS System, which is designed as a subsystem according to characteristic of DiffServ networks and the traffic measuring requirement. This subsystem has two function modules, per hop traffic analysis and end to end QoS performance measuring. The flow sampling technology, topology information collection and topo matrix based traffic calculation are involved in the per hop traffic analysis, to provide per link bandwidth utilization measuring for each traffic class and bottleneck and redundancy detection. The end to end QoS performance measuring introduces filter based flow information collection and flow pair matching method, and provide end to end delay and jitter measurement. Meanwhile, the flow deployment method is analyzed, considering the performance impact on the network devices to enable flow monitoring process. In the aspect of system realization, we gives the overall statements for the models design and data flow of whole system, and the method for each model.A congestion situation is simulated on the core routers by inpouring huge traffic through the backbone to verify the functions of the system. This system provides the detailed traffic information on the backbone router where congestion happens, and the bandwidth resource status of the whole DiffServ domain. The congestion chart in long time range gives the bandwidth utilization trend lines for per hop devices. In the end to end QoS performance testing, the delay, jitter and packet loss rate for the services of different QoS rank is given. The last part gives the sum-up and the disadvantage and improvement for this system.