光纖封包交換系統交換機的高服務品質排程

Abstract

由於各種類型的應用媒體不斷的增加，使得現代網路需求大量的頻寬，而且隨著不同的應用媒體有不同的網路連線品質需求，因此全光封包網路成為下一世代高速骨幹網路的首選。為了有效動態分配頻寬以達到各類型服務品質保證，並且能在非常短的時間內完成計算以達到real-time的要求，如何設計一具有高效率排程技術的全光封包交換機成為最關鍵的技術。 此論文架構於之前所提出的光分波多工全光封包交換機(QOPS)，我們提出了一個利用Hopfield類神經網路為核心模型的封包排程技術。透過問題的轉化，我們成功的將封包排程問題model成最佳化的問題。因為Hopfield類神經網路具有快速平行計算的特質，預期此模型的建立將能達到高效率封包排程的計算能力。

With the rapid growing demand of Internet bandwidth and the Quality-of-Service (QoS) requirement for various multimedia traffic, the Optical Packet Switching (OPS) technique which uses optical wavelength division multiplexing (WDM) become the future-proof solution for metro backbone network. In order to assure efficient bandwidth allocation and various QoS guarantees in a real-time fashion, the design and implementation of a highly efficient scheduling mechanism becomes crucial. In paper, based on the previously proposed 10-Gb/s almost-all-optical packet switching system (QOPS) for metro WDM networks, we propose a Hopfield Neural Network (HNN) based QoS-enabled optical packet scheduling algorithm which inherit the highly efficient property from the parallel computation nature of HNN.