透過您的圖書館登入
IP:3.143.22.26
  • 學位論文

10Gbps線速率八埠可擴充交換機之硬體實作

Implementation of Scalable 8-Port NTU-V Switch with 10Gbps Line Rate

指導教授 : 吳靜雄

摘要


隨著網際網路成為人們日常生活中大量資訊的重要來源,而且隨選視訊、視訊會議、即時通訊等諸多高網路頻寬需求的應用日益普及,對於網路的頻寬要求俱增。區域網路頻寬也已在近幾年內跨入百億位元乙太網路(10Gigabit Ethernet)的紀元,將在市區網路(MAN)以及大範圍網路(WAN)扮演重要角色。 由於網路內流量不斷的增加,網路節點的封包交換變得越來越重要,唯有提昇節點交換機的頻寬容量,才能使得網路系統的節點不再成為網路系統的流量瓶頸所在。除了要有每秒百億位元的網路界面卡之外,網路節點上的交換機(Switch)與路由器(router)的速度同樣也要進步到每秒百億位元等級,因此,交換機和路由器的研發顯得更加的重要。 常見的交換機種類有:輸入佇列、輸出佇列、虛擬輸出佇列以及結合輸入輸出佇列。輸入佇列交換機很容易遇到排頭阻塞(Head of Line Blocking)效應,理論計算顯示流通率只能夠達到0.586;輸出佇列交換機如共享記憶體交換機,利用複雜的演算法來進行交換,交換頻寬遠低於記憶體頻寬;而結合輸入輸出佇列交換機可以在對記憶體頻寬要求較低的前提下,得到相同的流通量。 本篇論文主旨是在於我們用電路實現了一個用於高速網路的高效能交換機,並將其命名為NTU-V交換機。在NTU系列交換機中,我們利用多層閂式交換平面的架構,擁有能直接交換不定長度封包的能力,同時硬體複雜度低,並且採用模組化設計方便系統擴充。構成交換機各級的次模組(sub-module)包含了:預先處理器(preprocessor)、時序器(sequencer)、交換平面(switch plane),以及記憶體管理單元(MMU),使封包在每一級間正確且有效地被傳送。 相較於前一代的NTU-IV交換機十六埠的架構,NTU-V交換機採用可擴充性的八埠架構,並以將資料匯流排(data bus)的寬度從原先的六十四位元減半為三十二位元,因此系統作用在比原先運作頻率(operating frequency)快兩倍的312.5MHz下。由於原先用於NTU-IV的0.18微米製程已無法支援在如此高的頻率下運作,所以NTU-V交換機在電路設計方面,採用台灣積體電路製造公司(TSMC) 0.13微米(um)製程的標準元件式(Cell-based)超大積體電路(VLSI)技術來實現。 我們藉著減半的資料匯流排寬度和0.13微米製程高密度的優勢下,大幅度的縮減次交換機(sub-switch)的單元面積。解決原先十六埠NTU-IV交換機因為面積過大而影響它的實作以及應用的限制。可擴充的八埠NTU-V交換機不僅在次交換機的單位面積遠小於NTU-IV次交換機,在擴充成為十六埠的規模之下,總面積也是優於NTU-IV交換機而利於實現。這樣的標準或許可以使用在像是兆位元交換機(Terabit Switch)這種大規模交換系統的架構上,可見NTU-V交換機在網際網路上具有相當的發展性。

並列摘要


In this thesis, we propose and implement a novel high speed switch named NTU-V switch which is based on multi-plane cross-bar switch fabric. This switch has low complexity and the ability to handle variable length packets asynchronously. In addition, our switch will also support the multicast. The concept of modular design is embedded into the architecture of NTU-V Switch. Therefore NTU-V Switch has great scalability in port count and line-rate. Simulation results show that this switching architecture can eliminate head-of-line blocking (HOL blocking) effect. This switch design can emulate the performance of output queue (OQ) switch without complicated contention resolution algorithm and arbiter. NTU-V Switch is composed of preprocessors, sequencers, switch planes, and memory management units. Each component has the ability of self-routing. We introduce the architecture from the point of hardware design. Moreover, we use the 0.13 um cell based design flow to implement an 8x8 NTU-V Switch prototype with line rate at 10Gbps. The prototype may be used as the building block of a large switching system, e.g., Terabit Switch. Our design flow includes behavior level coding (Verilog), logic synthesis (Design Compiler), and place and route (Astro). It shows that the switch has low delay and almost 100% throughput under uniform traffic condition.

參考文獻


[1] T.Y. Wang, K.T. Chen, H.P. Shiang, M.L. Yang, H.W. Tsao, and J. Wu, “A novel high speed asynchronous scalable variable-length self-routing packet switch,” APCC Sept. 2003, vol. 3, pp. 1181-1189.
[2] S. Iyer, R. Zhang, and N. McKeown, "Routers with a single stage of buffering," ACM SIGCOMM Aug. 2002, Pittsburgh, USA.
[3] M.G. Hluchyj, and M.J. Karol, “Queueing in high-performance packet switching,” IEEE J. Select. Areas Commun., vol. 6 no. 9, pp. 1587-1597, Dec. 1988.
[4] M.J. Karol, M.G. Hluchyj, and S.P. Morgan, “Input versus output queueing on a space-division packet switch,” IEEE Trans. Commun., vol. Com-35, pp. 1347-1356, no. 12, Dec. 1987.
[5] N. McKeown, A. Mekkittikul, V. Anatharam, and J. Walrand, “Achieving 100% throughput in an input-queued switch,” in IEEE Trans. Commun., vol. 47, no. 8, pp. 1260-1267, Aug. 1999.

延伸閱讀