全光都會網狀/環狀骨幹及其互聯之光纖及無線擷取網路整合技術---子計畫二：具QoS保證機制都會光纖網路資源控制策略設計之研究

Abstract

在這三年期計畫中，我們的主要目標是對全光WDM 服務品質保證都會網路中的光群聚 交換(Optical Burst Switching)資源保留排程演算法進行設計OBS Node 所連接到的光纖都會 環狀網路(Optical Metro Ring Network)作unicast 和multicast 訊務彙集方法的研究設計及分 析，以及還有FTTH 光纖到家技術中EPON/GPON 的DBA 方法的研究。近幾年，由於高密 度分光波長多工技術 DWDM 不斷的進步，都會光纖網路的頻寬戲劇性的增高，讓使用 DWDM 的都會光纖網路技術悄然成形。而在光交換網路技術的研究方向上，OBS 受到很大 的青睞，相對於光路徑交換(OCS)，OBS 可以提供相當高的頻寬使用率，同時亦可以在無須 光隨機存取記憶體的需求下，達到封包遞送的功能。換句話說，OBS 能同時擷取OCS 以及 OPS 之所長並且避開OCS 以及 OPS 的缺陷。但是隨著網路使用者的需求，如何在OBS 網 路內作到提供有效率的頻寬分配與不同等級的網路服務品質(QoS)保證技術仍亟需解決。基 本上，在OBS 的部分，第一年我們將進行深入的文獻研讀及資源保留排程演算法的設計； 第二年則更進一步的研究設計新的OBS 通訊協定來解決現有通訊協定的問題。最後第三年 將對前兩年所發展的OBS 資源保留排程演算法及通訊協定做數學理論分析並加以改善。同 時同步光纖(SONET/SDH)環狀網路被大幅應用在都會光纖網路的基礎建設上。隨著光纖網路 發展的進步，訊務流量激增。並且由於分波多工(WDM)技術成熟及應用下，光纖網路的傳輸 容量大幅的提升。但相對於一個波長提供的頻寬，訊務的流量仍是相當的小。所以，可以利 用訊務彙集的技術有效的利用資源。但是考量因應網路電視等應用產生的multicast 訊務，就 會使的訊務彙集技術變的更加的複雜了。而在訊務彙集部份，第一年先針對動靜態unicast 訊務進行訊務彙集方法研究；第二年則是對multicast 做訊務彙集方法的研究。最後第三年將 對前兩年所發展的unicast 和multicast 務彙集演算法及通訊協定做數學理論分析並加以改 善。而FTTH 用以提供比ADSL 更快速，頻?更高的上網服務，且多家國際研究諮詢機構， 均認為2006 年～2008 年是FTTH 技術快速應用和推廣的幾年。而在EPON/GPON 等技術中， 如何藉由DBA 的方法同時滿足各收費服務用戶QoS 的需求下，並顧及公平性問題依舊值得 探討。在此部份，第一年深入研究設計EPON/GPON 中的DBA 方法；第二年DBA 則採類神 經網路預測器的功能來提升性能的表現。

In this three-year project, our main objectives are the design and analyses of the OBS resource reservation algorithm in the QoS-guaranteed metro optical network, of the uni-cast and multicast traffic grooming in the metro optical ring network which connects the backbone by an OBS node, and of the DBA algorithm in EPON or GPON. In recent years, since the bandwidth of the metro optical network is rapidly increasing according to the Dense Wavelength-Division Multiplexing (DWDM) swift development, the metro optical network with DWDM equipment would come true. And optical burst switching (OBS) plays an important role in the optical network switching technology. The OBS provide improvements over wavelength routing in terms of bandwidth efficiency comparing with optical circuit switching (OCS) and transmits the burst without the memory contrast to optical packet switching (OCS). In other words, OBS combine the best of optical circuit and packet switching while avoiding their shortcomings. However, how to provide the QoS in the OBS network still has no proper solution. In principle, while we give the decent study and preliminary design in the first year, we would design a new OBS protocol to solve or improve the drawback of the present OBS protocols in the second year at this section. Finally, we would improve and mathematically analyze the OBS resource reservation scheme and OBS protocol which are designed in the previous two years in last year. In the meanwhile, synchronous optical network (SONET) ring networks have been widely deployed for the optical network infrastructure. The progresses of optical networks evolve with time; meanwhile, the carried traffic streams surge. The transmission capacity of an optical network largely increases because of the development and application of wavelength division multiplexing (WDM) technologies. Nevertheless the required bandwidth of a traffic stream is also much smaller than the bandwidth capacity of a wavelength. However, if we additionally consider the multicast traffic which is the result of the video online, then the traffic grooming technology would become more complexity. In this section, as we perform the design of the nui-cast traffic grooming for static or dynamic traffic in the first year, the multicast traffic grooming for static or dynamic traffic would be developed in the second year. Finally, we would improve and mathematically analyze the uni-cast and multicast traffic grooming under static or dynamic traffic model in last year. Fiber to the home (FTTH) is used to provide the higher bandwidth and more quickly connect to the network than ADSL. Many authorities think that it would widely develop and popularize from 2006 to 2008. Hence, how to attend to the QoS and the fairness by a dynamic bandwidth allocation algorithm it is still worth to research in EPON or GPON. Therefore, we would enhance the improvement of the DBA algorithm in the first year and meantime we perform the better performance by adding a neural network predictor in the second year.