Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/19313
標題: 利用光纖緩衝器排程法於回饋式非同步光封包交換器之效能評估
Performance Evaluation of Asynchronous Optical Packet Switches with FDL Buffers
作者: 陳均銘
Chen, Jyun-Ming
關鍵字: Asynchronous optical packet switch
非同步光封包交換器
Fiber delay line
Feedback buffering
FDL buffer management
光纖延遲線
回饋式緩衝器架構
光纖延遲線式緩衝器排程法
出版社: 資訊科學與工程學系所
摘要: 在網路交換器技術中,解決封包之間的衝突問題向來是重要的課題之一。當多個封包同時要通過同一個輸出口時就可能會發生訊號碰撞而產生干擾;為了避免碰撞,使用緩衝器(buffer)來調整先後順序是常用的做法,然而光封包交換器(optical packet switching,簡稱OPS)受限於技術,尚未有如電子式隨機存取 (ram-like) 的緩衝器能利用,因此使用光纖延遲線(fiber delay line,簡稱FDL)來延遲光訊號以達到緩衝目的,就成為一種簡單且有效率的方法。 本篇論文的主要目標在於研究光交換器架構的效能和應用;在過去研究光封包交換器上,非同步式 (asynchronous) 光封包交換器是最能切合實際網路主流環境:Ethernet-like IP-traffic架構,早期研究非同步光封包交換器是以前饋式 (feed-forward) 架構為主,到目前此架構已有不少精確的數值分析方法被提出,然而這類架構為了增強其緩衝區深度,都脫離不了大量FDL需求;另一類架構則是使用回饋式 (feedback),這類架構允許FDL可從時間上的串聯,使訊號能以回繞的方式有效提昇的緩衝區的使用效率,也因此間接地降低FDL的需求量,然而有效的數值分析方法至今仍未有成熟的成果。 因此在本篇論文,我們希望能先以程式模擬的方式對非同步回饋式光交換器的效能做較深入的實驗和探討,以期建立起未來發展數值分析方法的基石;整篇研究的論題分成三個部分:首先我們比較前饋式和回饋式在緩衝區大小需求和緩衝區使用效率上的差異,作為回饋式優於前饋式架構的理論依據。而後,我們在回饋式架構設計兩種取向相反的預留型 (reservation) 封包排程法做對比,利用這兩種排程法交互調控回饋式架構裡兩項重要的影響參數:『緩衝區大小』和『訊號回繞次數限制』,觀察系統的效能並比較對各別兩種排程機制的影響性。 最後,由實驗觀察中我們發現透過兩項參數調配下,最低封包掉落率的分布趨勢會有一些特殊的現象,由結果我們做出數項推論:1. 單獨將一參數提高 (另一項參數值固定),並不會無限地增加系統的最佳封包遞送效能,反而在適當的參數配置比例才能達到的效能最佳化。2. 在只單獨變動『訊號回繞次數限制』的例子中,增加『訊號回繞次數限制』會使效能提昇但最終會到達極限值。3. 在只單獨變動『緩衝區大小』的例子裡,增加『緩衝區大小』則發現在效能到達預期的極限前,會先於某一點出現最低封包掉落率的特殊現象。對這些結果我們在論文中皆做了相當程度的驗證和討論。
Resolving packet contention is a main issue in the design of network and switch architecture. Packet contention in network switches occurs when multiple packets are destined to the same output port at the same time. Buffering is a conventional solution for contention avoidance. In the optical domain, buffering typically uses fiber delay lines (FDLs) to delay an optical signal for a discrete, fixed amount of time. Many asynchronous optical packet switches (OPSs) rely on the use of feed-forward buffers. However, they often require a great number of FDLs for buffering contending packets, and they become economically infeasible for implementation. The feedback buffer is a promising approach to reducing FDL buffer size, because it can efficiently circulate packets through an FDL buffer to achieve a prescribed amount of delay time. The main objective of this thesis is to study the performance of asynchronous OPSs with feedback FDL buffers. We evaluate the performance of feedback FDL buffers, and compare it with that of feed-forward buffers with respect to buffer size and buffer utilization. Two packet forwarding schemes are proposed for scheduling packets and managing FDL buffers, respectively. We examined two major system parameters of the feedback architecture: they are FDL buffer size and recirculation limit. Through simulation, we analyze their joint effects on system performance. Our simulation results show that raising recirculation limit can effectively enhance the system performance. The results indicate that through the proposed buffer management scheme, we can determine an optimal buffer size and achieve the lowest packet blocking under various working conditions.
URI: http://hdl.handle.net/11455/19313
Appears in Collections:資訊科學與工程學系所

文件中的檔案:

取得全文請前往華藝線上圖書館



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.