【作者】 陈文涛；

【导师】 曾烈光；

【作者基本信息】 清华大学 ， 信息与通信工程， 2009， 博士

【摘要】 以太网在城域网中的应用,即城域以太网,是目前的研究热点,且受到业界的广泛关注。在诸多城域以太网解决方案中,全以太网的解决方案由于能够最大程度地继承以太网价廉物美的特点,被认为是最有竞争力的方案之一,但是其故障恢复和QoS方面的性能还有待提高;而在TDM网络大量铺设的现状下,基于TDM设施的解决方案是一种很实用的过渡性方案。因此,本文主要研究全以太网解决方案的故障恢复和QoS问题,以及基于现有TDM设施的解决方案中的电路设计问题。论文的主要创新成果和结论如下:1.针对全以太网的解决方案,提出了一种基于自保护生成树的快速故障恢复方案,能够实现对任意单故障的快速恢复。从理论角度证明了,该方案对生成树数目和网络拓扑的要求都达到了下限。此外,仿真结果表明,故障恢复时间为几十毫秒量级,满足城域以太网的需求。2.针对全以太网解决方案的拥塞控制问题,设计了一种高精度低复杂度的速率控制电路。针对硬件实现,对漏桶算法进行改进后,所设计的速率控制电路具有如下优点:发送速率的调整精度高,满足IEEE 802.3ar工作组所提出的高于1%的需求;电路简单,规模小于以太网MAC电路的5%。3.针对全以太网的解决方案,提出了一种用于降低故障发生后平均延时的分布式切换方法。理论证明和仿真结果表明,分布式切换在不恶化工作树平均延时的前提下,能够明显减小网络的单链路故障下平均延时。4.针对基于现有TDM设施的解决方案,提出了一种基于帧间插的以太网到多路E1反向复接器设计方案并用FPGA进行了实现。相比已有的字节间插方案而言,帧间插方案的优点是,只要有一路E1正常工作就能实现部分数据帧的传输,从而可以用网管功能实现快速自动的故障隔离。5.从电路设计的角度,提出了一种利用异步采样电路的不确定性提高FPGA的设计安全性的方案。异步采样电路的特点是,每次上电后的输出序列都不一样,而且和温度等外界因素有关。这种不确定性将使得剽窃者每次采样到的配置数据和验证数据几乎都不一样,增加了剽窃的难度。更多还原

【Abstract】 Ethernet has been the most dominant networking solution in local area networks, and it is now expanding to the Metro Area Networks (MAN). The application of Ethernet in the MAN, named metro Ethernet, is a hot topic, which is also widely concerned by the industries. There are various metro Ethernet solutions. The native Ethernet solution is considered as one of the most competitive solutions because of its low cost. However, it faces resilience and Quality of Service (QoS) problems. Moreover, the TDM-based solution is a practical temporary solution, because the TDM infrastructures have been wildly deployed. Therefore, we mainly focus on the resilience and QoS issues in the native Ethernet solution, and circuit design issues in the TDM-based solution in this thesis. The main achievements and contributions are as follows.1. We propose a self-protected spanning tree based recovery scheme for the native Ethernet solution to protect against any single failure. It is theoretically proved that the scheme’s requirements to both the number of the trees and the network topology reach the lowest bound. Moreover, simulation results show that the recovery time is about several tens of milliseconds, which meets the metro Ethernet’s resilience requirement.2. We propose a fine-adjustable low-complexity rate control circuit for congestion control in the native Ethernet solution. Based on the hardware implementation oriented modifications to the token bucket algorithm, the designed rate control circuit has the following features. The rate reduction is fine-adjustable, which meets the 1% or better granularity requirement presented by IEEE 802.3ar task force. Moreover, the circuit is simple, and its area is less than 5% of the Ethernet MAC.3. We propose a distributed switching mechanism to reduce the average delay after link failure for the native Ethernet solution. Both theoretical proofs and simulation results show that the distributed switching mechanism is able to reduce the average delay after single link failure without affecting the original average delay of the working tree.4. We propose a frame granularity scheme for Ethernet over E1, which is one of the TDM-based solutions, and implement it with FPGA. Compared with the byte granularity scheme, the advantage of the frame granularity scheme is that even if only one E1 line works normally, some frames can still be tranmitted to the far end. Thus, the failed E1 links can be isolated using the network management function.5. We propose an asynchronous sampling circuit to secure the configuration of FPGA. The feature of the asynchronous sampling circuit is that its output sequence is different every time the board is powered on, and the sequence is even related with the outside conditions such as air temperature. This uncertainty prevents the possibility of cracking the system by copying the sequence.更多还原