Traffic Schedule And Control Technology Research On Distributed Network

With the rapid development of modern information network technology, users’ demand for the network service capacity is also increasing.Due to constraints of network bandwidth, network facilities and other factors, how to improve the efficiency of network service have become an urgent problem of modern network.As a typical distributed network, Content Delivery Network enables users access the content nearby, which greatly improves the efficiency of services.The load balancing in CDN is the key part in network system and the decisive factor of CDN performance. Load balancing provides an effective method to enhance the data processing ability and improve network availability based on original network. It distributes the network traffic based on load capacity of the cache server.Therefore a stable and reliable load balancing algorithm has important application value and practical significance.This thesis divides load balancing in CDN into two parts, including the network traffic prediction and network traffic scheduling combined with the prediction, which will improve the efficiency of load balancing. This thesis summarizes various prediction algorithms, including ARMA model, ARIMA model, exponential smoothing model and kalman fiter prediction model and traditional CDN network load balancing algorithm, such as round robin algorithm, weighted round robin algorithm, minimum connection scheduling algorithm. On this basis, for the first part, a distributed traffic prediction scheme based on intelligent optimization is studied and implement. In this scheme, Fruit fly optimization algorithm is adopted to optimize smoothing coefficient of the traditional triple exponential smoothing forecasting model, which effectively improves the efficiency of traffic prediction. For the second part, a network traffic scheduling scheme is studied and implement. Due to the distributed network traffic scheduling model is a multi-dimensional knapsack problem, fruit fly optimization algorithm is adopted to solve this scheduling problem for the realization of load balancing.Finally, by testing two schemes in simulation experiment, the results show that the traffic prediction scheme can improve the efficiency of prediction and the scheduling scheme can ensure load balancing in consideration of network comprehensive information.