A Load Balancing Algorithm Based On Neural Network Prediction Model In Cognitive Networks

With the rapid development of Internet, the communication demands among people are increasing and the existing internet structure cannot meet such various requirements. Therefore, it is necessary to design a new generation of intelligent network. The theory based on the issue named "The cognitive network QoS technology based on Cognitive model of network behavior" which was National High Technology Research and Development Program 863. Cognitive Technology is a burgeoning concept. In the network, the terminal distribution, random movement features, various QoS requirements of users will cause uneven distribution of network traffic, local business overload, congestion of heavy loaded areas, leading to the increase of packet loss and delay operations, while in the light loaded areas, the idle resources are not fully utilized.Self-aware network-specific learning and reconfiguration capabilities of cognitive networks make the load balancing more effective. One of the most effective ways to improve the utilization of network resources is to make the request data scheduled reasonable by load distribution and traffic scheduling in order to achieve the best treatment ability of the network systems.Based on the research of traffic scheduling algorithms and the traffic prediction model, the essay proposes an adaptive scheduling algorithm based on the traffic prediction model on the cognitive networks NNPMA(n adaptive load balancing algorithm based on neural network prediction model). According to the results of network traffic forecast and the changes with network load conditions, the business flow is controlled in real-time, network request is effectively dispatched to all network nodes. This effectively solve the problem that some servers are overloaded while some servers are idle, so the requests for each node are allocated evenly across the network, reducing network congestion, ensure the business QoS requirements. Finally OPNET is used to emulate the NNPMA process to prove its effectiveness. Comprehensive simulations show that NNPMA is more effective on load balancing, and it does not produce a large network cost.