【作者】 周榜兰；

【导师】 花欣；

【作者基本信息】 西南交通大学 ， 计算机应用技术， 2008， 硕士

【摘要】 随着Internet上业务流量的快速增长,网络应用的多元化发展,传统IP网络提供的尽力而为服务已经不能满足承载实时业务的需求。如何在故障发生时保证网络的连续性和服务质量(QoS),保持高水准的业务可用性,使传输网络能够自动检测失效并从故障中恢复成为当前迫切需要解决的问题。作为未来骨干网的核心技术,多协议标签交换(MPLS)技术通过标签交换机制,不仅可以提供比传统IP更有效的QoS保证和流量工程,也具有很强的网络生存能力。MPLS网络能够在节点或链路失效后,不需要人为的干预,就能自动的恢复受影响的业务,使整个网络保持健壮性和高效性。MPLS网络自愈恢复技术从故障检测技术和故障恢复技术两方面保证了MPLS网络的可用性和稳定性。目前基于IP路由协议的恢复机制至少需要几秒到几分钟的时间,这将导致大量分组的丢失,造成严重的服务质量问题以及网络性能的下降。传统IP网络对实时应用(如语音)进行准确故障检测方面的能力有限,并且不具备秒以下的间歇性故障修复功能。伴随着VoIP应用的激增,实现快速网络故障检测和修复越发显得必要。MPLS网络自愈恢复技术能够在故障发生后,提供比IP层更快的反应时间,可以快速的检测到网络故障并从故障中恢复过来,维持网络的服务连续性。本文对MPLS网络故障检测技术进行深入的研究,采用双向转发检测BFD作为MPLS网络的故障检测技术,详细的设计和实现MPLS BFD技术,并将MPLS BFD技术应用于多核网络设备,使MPLS网络故障检测时间达到毫秒级,大大缩短了MPLS自愈恢复的故障检测时间。MPLS网络检测到故障后,需要有一个较好的方案进行故障恢复。目前比较著名的两种MPLS故障方案有Makam方案,Haskin方案。Makam方案的优点是几乎没有分组的重排序问题,但却有分组丢失和保护时间较长的缺点。Haskin方案提供快速的路径切换,但是当工作路径上故障清除后,在流量从备份路径切换回原工作路径时,数据分组的重排序问题比较严重。因此如何设计出一种新的故障恢复方案,使得故障恢复速度快,报文不丢失,并且报文不发生重排序仍然是目前研究的热点。本文在对现有故障恢复方案研究的基础上,提出了一种新的故障恢复方案——基于BFD的故障恢复方案,新方案利用局部恢复机制,建立备份路径,当故障发生后流量快速切换到备份路径,减少了由于流量切换而造成的报文丢失及报文延迟。针对报文失序的问题,本方案提出了一种新的解决方法——利用失序控制标签实现了对流量顺序地控制,避免了流量切换带来的报文失序。经过组网测试,新方案具有较快的故障恢复速度,对报文丢失,报文失序现象都有明显地改善效果。更多还原

【Abstract】 The traditional IP network now fails to serve the need of bearer real-time service in an effective way with its best-effort service against the backdrop of ever-increasing Internet service traffic and gradually diversified network service. There is an urgent need to guarantee the network continuity and Quality-of-Service (QoS) against network fault in order to obtain ideal service availability to help the automatic detection of network failure and recovery from that failure. As the core technology for the future backbone network, Multi-Protocol Label Switching (MPLS), by means of label-swapping technique, is expected to offer the Quality-of-Service (QoS) and traffic engineering more effective and potential than ever before. The MPLS network automatically resumes the interrupted service and maintains the robustness and efficiency of the network as a whole even in the case of node failure or link failure. The self-recovery technology of MPLS network is used to protect the network feasibility and stability in terms of fault detection technique and fault recovery.It takes at least several seconds or even a couple of minutes for the current recovery mechanism of the IP-based routing protocol, thus leading to a great deal of packet loss, the bad quality of service and the degradation of network performance. Due to the traditional IP network’s undesirable capability of accurate fault detection for real-time applications like voice and for lack of intermittent recovery ability below second level, it is imperative to put fault detection and repairment for network into effect in accordance with the expansion of VoIP application. The self-recovery technology of MPLS network can, in response to the network fault, offer a reaction time less than IP layer and quickly detect the network fault and recover from that fault to remain the continuity of the network service. This paper, on the basis of in-depth discussion on the fault detection technique of the MPLS network, significantly reduces the detecting time of MPLS self-recovery to millisecond level by adopting Bidirectional Forwarding Detection (BFD) as the fault detection technique of MPLS network, and by careful design and realization of MPLS BFD as well as its application into multi-core router.There needs to be a better scheme for fault recovery after the network is detected by MPLS. Of all the recovery schemes, there are two most famous ones, namely Makam and Haskin. Makam almost avoids the problem of packet reorder but suffers packets loss and relatively long protection time, while Haskin provides quick path switch but suffers severe problem of packet reorder as soon as the network traffic is redirected into the original working path to backup path after successful repair on working path. Thus, it has become the key point on how to develop a new type of fault recovery scheme so as to speed up the fault recovery while avoiding packet loss and packets disorder.This paper offers a new kind of fault recovery scheme, namely the BFD-based fault recovery scheme on the basis of the research into the existing recovery schemes. The BFD-based fault recovery scheme possesses local repair mechanism to set up backup path so that the network traffic is quickly redirected into backup path once fault occurs, thus effectively avoiding packet loss and packet delay caused by traffic switch. This paper presents a new type of solution to the problem of packet disorder, namely utilizing disorder-control-label to control packets order and to avoid packet disorder. Tested in a real experimental network, the new type of fault recovery scheme is proved to be quick in fault recovery and effective in avoiding packet loss and packet delay.更多还原