Research Of Network Traffic Measurement And Load Balancing Technology

With the rapid development of computer network technology, network scale is becoming larger and more complex, Web applications become increasingly complex and diverse, Network traffic monitoring is a very important part of computer network application. As a part of Network Management, which play a very important role in network performance analyses, anomaly detection, state monitoring, capacity planning and load balancing.This paper discusses and compares the popular Network traffic acquisition technologies, and puts forward Overall Framework based on NetFlow network traffic monitoring system linked with the demands of network integrated management. First, the paper introduces the system architecture and functional modules, and then makes separately detailed design of system modules which including four modules: flow collection, traffic statistical analysis, data storage and management and flow indication achieves features such as real-time display of flows, statistical classification of flows, link load sequencing and decision of large flow. Finally, the paper gathers the real network flow data from a campus network environment for the experimental, tests the functionality of the system, and analyzes the performance of some of the functions.According to the network traffic link load unbalancing and traditional algorithm for dynamic load distribution algorithm lacking time-validity, which easily leads to flow fluctuations and the problem of inefficient, this paper proposes a Load Distribution Algorithm based on the flow classification (AFDA). This new hybrid approach that divided a flow into the large flow and the small flow, which performs dynamic load balance of long-lived flows (large flows), while forwarding short-lived flows (small flows) statically, to achieve the efficiency of the load schedule. This algorithm is evaluated by simulation with NLANR traces, which shows that flow flapping and load-balancing performance is considerably decreased that significantly outperforms traditional schemes.In summary,our work presents a useful network traffic monitoring application platform and solutions to several key problems about the load distribution of the flow,and has academic and practical value for further research.