雙網路/對等節點架構之分波多工網路

Abstract

在這篇論文中，我們介紹一新型光纖網路架構，並根據其特性將之命名為雙網路/對等節點架構 (DNTN)。此架構的概念源於將現有許多獨立的小型區域網路 (LAN)或中型都會網路 (MAN)，連結成一個有效率且高可靠度的大型廣域網路(WAN)。 一般而言，這些區域網路或都會網路各有其不同之需求，可能沒有相同的網路架構及封包傳遞協定。但藉由雙網路/對等節點架構，我們可以簡單而有效率地將已有的區域網路或都會網路，串連成大型廣域網路，並且對原有各網路運作不會產生明顯干擾。 許多常見的網路結構，如RING、STAR、perfect Shuffle及Manhattan street網路等，都可以在DNTN架構中被採用。在這些不同的組合中，我們發現到有些傳統網路以DNTN架構組合起來後，具有高可靠度，低平均傳送距離，及簡單封包傳遞協定等優點。輻射型網路(Radiation Network)便是我們所提出，一種以DNTN架構之方法，結合環型網路(Ring Network)及星型網路(Star Network)而成之新型網路。 從本論文的分析結果中，我們瞭解DNTN架構可以選用許多不同的封包傳遞協定，來改善整個系統的等待時間(system waiting time)，以提高網路的效率。而且由於DNTN架構帶來波長重用(wavelength reuse)的特性，可以使網路節點具有更高的資料處理能力。這些優點，使得我們所提出的輻射型網路，非常適合波長分工的廣域光纖網路之用。我們將在論文中仔細討論波長分工規劃下的輻射型網路，呈現其互補環型網路及星型網路之優缺點的特性，說明依DNTN架構概念所建構而成的輻射型網路，確實是非常適合未來光纖網路規劃的新型結構。

In this paper, we present a novel approach dedicated for planning optical networks, the Dual Networks Twin Nodes (DNTN) architecture. It originates from the consideration to efficiently and reliably connect many existing, blooming, ever independent but now closely correlated local area networks (LAN) or metropolitan area networks (MAN). To meet the individual requirements, these LAN/MAN do not always have the same topologies and routing algorithms. By adopting the proposed DNTN architecture, we may simply connect these existing LAN/MAN into a wide area network (WAN) with little interference on the existing network operation. Many familiar network topologies, such as ring, star, perfect shuffle and Manhattan street networks etc., could be adopted in the DNTN structure. We find that some excellent combinations can achieve high reliability, low mean hop distance as well as simple routing algorithm. We further propose a Radiation network constructed from a ring and a star subnetwork, being a novel network topology in the sense of DNTN architecture. A wavelength-division multiplexing (WDM) based Radiation network is studied in this paper. The Radiation network shows the merits that each subnetwork actually takes advantage of the other's special feature to compensate its own weakness. Thus the construction turns out to be perfect combination of ring and star networks. From the analysis results, we see that alternative routing algorithms can be employed to improve the system waiting time. Every node can also have higher service capability because of the possibility of wavelength reuse in the construction. Especially, the proposed Radiation network could be an excellent candidate for WDM-based wide area networks.