【摘要】 提出一种改进的ATM交换机多输入队列的神经网网络调度算法(IMIQM),其调度策略采用每条入线在同一时隙内可传送多于1个信元的策略,并提出一种用于Hopfield神经网络(HNN)控制信元调度的新的能量函数。利用计算机进行仿真模拟,在业务流模型和负荷相同的情况下,当交换机规模N为150、多输入队列(开窗数)为5时,IMIQM的最大吞吐率可以达到0.904,普通的多输入队列法(MIQM)为0.856,而窗口方法(WM)为0.886。结果表明,IMIQM与MIQM、WM相比提高了吞吐率,或在吞吐率相同的情况下IMIQM更容易用光电子技术实现,且由于HNN的高度并行的数据处理能力,能够实现大规模交换结构的实时调度。更多还原

【Abstract】 An improved multiple input-queuing method(IMIQM) used in asynchronous transfer mode(ATM) switch by a Hopfield neural network(HNN) scheduling algorithm is proposed.The scheduling policy of more than one cell transferred in each input line during every time slot is employed in the IMIQM,and a new energy function is presented to accomplish this policy in HNN model.The computer simulations show that the maximum throughput is up to 0.904 of IMIQM when the switch scale N is 150 and the number of queues is 5,but it is only 0.856 and 0.886 for general multiple input-queuing method(MIQM) and window method(WM) respectively with the same scale and queue(window) number under the same traffic model and load.It means that the throughput of IMIQM is improved greatly compared with the MIQM and WM.The IMIQM,therefore,can be used in real-time optimization scheduling of large-scale ATM switches.更多还原