A Cost-Effective Protection and Restoration Mechanism for Ethernet- Based Networks: an Experiment Report Yi Lei Chung-Horng Lung Anand Srinivasan Department of Systems and Computer Engineering EION Inc. Carleton University, Ottawa, Canada Ottawa, Canada Abstract Protection and restoration at the physical layer is fast, but may require dedicated hardware. On the other hand, IP, although it does not rely on specific topology, is slow. Protection based on MPLS is effective for rerouting traffic to a pre-established backup LSP in the case of failures. Failure detection plays a dominant role in protection and restoration. This research presents an approach for addressing MPLS link and node protection by making use of auto-negotiation for Ethernet. The method is efficient to detect failures in an Ethernet-based network and can be easily adopted to support MPLS protection with low overhead. The experiment results show that traffic recovery can be achieved in sub-30ms. Keywords -- MPLS, link/node protection, restoration, Ethernet, auto-negotiation 1. Introduction Protection against and restoration after link and node failures can be performed at several layers in a network [5]. The mechanisms for protection and restoration can be placed at the network layer (IP layer), data link, or the physical layer. Protection/restoration at the network layer uses IGP protocols [1] such as OSPF, ISIS and RIP to maintain and update its routing table. In the case of a failure, the IGP protocol takes into account the topology change and re-computes the routing table by using the shortest path algorithm. When all the routing tables in the network are recomputed, the traffic traveling through the original route will be redirected through the new route. Even though this has been proven to be robust and survivable, the recovery coverage of this methodology is slow and can take a long time, from several seconds to minutes [6]. Another approach is to provide protection/restoration in the physical layer, where SONET is used for optical transmission. Automatic Protection Switching protocol (APS) [3] is used to switch over from a failed fiber to a protected fiber. APS involves a window of time of 10 ms for fault detection and can achieve a recovery in 50 ms [5]. However, this fast recovery is achieved at the expense of inefficient use of bandwidth and is typically limited to SONET/SDH ring-based systems [1]. Protection at the physical layer is fast but may require dedicated hardware, such as SONET protection ring. In addition, physical layer alarms are not always available [9]. Conversely, IP does not rely on specific topology, but it is slow. Detection with RSVP hello -based method usually is also slower than layer-2 alarm-based approach. MPLS, which is between the IP and link layers, supports recovery mechanisms that provide a trade-off between recovery speed and deployment cost [7]. MPLS, allows a Label Switch Path (LSP) to be set up before the traffic arrives to support fast recovery. Ethernet is currently getting more popular due to its simplicity and lower cost. Ethernet is battling its way out of the enterprise networks and entering the metro areas or even the core of the network [8]. The Metro Ethernet Forum is supporting a technique that utilizes MPLS to enhance the resiliency of the LAN technology. One of the motions of the Metro Ethernet Forum is to develop the Ethernet-based metro Networks Protection for 50 ms restoration using MPLS [4]. This paper focuses on protection for Ethernet networks using MPLS. Not many actual reports on protection are available for Ethernet networks. The aim of the experimental research is to develop a lightweight, yet efficient method (<50ms) to effectively support MPLS protection and restoration. The method does not use signaling protocols, which introduce higher overhead and are more expensive. Rather, the approach takes advantage of a simple mechanism, auto-negotiation, designed for the Ethernet. This paper also presents concrete experiment results. The paper is organized as follows. Section 2 discusses MPLS protection and restoration. Section 3 presents experiment results. Section 4 is the conclusions. 2. MPLS Protection and Restoration In MPLS networks, several different methods have been proposed for detecting link failure. Basically, there are two categories for detecting failure: it can either be detected by the signaling protocols or by the physical layer. In the MPLS, RSVP-TE is the signaling protocol used to set up explicit LSP, and LDP is the protocol used to set up the hop-by-hop LSP dynamically. In both