【摘要】 对基于TCP协议的拥塞控制局限性和常见的拥塞控制算法进行分析,提出了一种全局性的拥塞控制策略处理新方法。通过网络监控软件获取网络设备的端口参数信息,利用神经元网络单层感知器学习算法检测网络拥塞发生,记忆当前网络状态的参数,再判断新输入模式是否导致拥塞发生,进而修改参数并避免网络拥塞。MATLAB仿真结果表明,使用这种神经元网络模型能够对网络拥塞进行有效地控制,提高网络传输性能。更多还原

【Abstract】 Based on TCP protocol congestion control limitations and the common characteristics of network congestion algorithm is analyzed, then a new way of congestion management with processing in policy is put forward and the happening of network congestion is tested. By means of network monitoring software, access to equipment’s port parameter information, then use of a learning algorithm in single-layer perceptron of neural networks, record the current state of network parameters, and then judge the new input pattern whether or not lead to cause the congestion occurred. Finally, the revised network’s parameter will avoid congestion. The MATLAB simulation results shows that with the use of neurarks we can make a effective control to the network congestion, thus enhances the capacity of network transportation.更多还原