Progress Report No. 15. Shared Segments Protection

Transcription

1 NEXT GENERATION NETWORK (NGN) AVAILABILITY & RESILIENCE RESEARCH Progress Report No. 15 Shared Segments Protection The University of Canterbury Team 18 April 2006

2 Abstract As a complement to the Canterbury NGN report 14 titled Service Availability and Recovery Time Models, this report introduces a promising network resilience mechanism named Shared Segments Protection (SSP), which can achieve the optimal trade-off solutions between connection availability, recovery time and spare capacity efficient. By dividing a primary working path into a sequence of segments and protecting each segment separately (or dividing the whole network into different domains where a sub path segment in one domain must be protected by the resources in the same domain), SSP can achieve a high scalability and fast recovery time for a modest sacrifice in resource utilization. 1. Background In the traditional IP networks with a connectionless transmission mode, data traffic on a link subject to a failure is recovered via Layer 3 or IP rerouting, or by some lower layer mechanisms such as SONET Automatic Protection Switching. These options may be used to restore IP packets of which the corresponding connections do not require further guarantee of service continuity and restoration time. With the emergence of some commercially important and delay-sensitive applications, such as high-priority voice and video traffic, equipping the connections of these applications with an end-to-end service-guaranteed protection and restoration plane becomes one of the major design issues in an NGN. The above task is particularly important when the photonic infrastructure is adopted in the backbone networks, where a fiber cut may influence million of end-users with different service requirements. In such a circumstance, not only the restorability of any service interruption must be guaranteed, but the restoration time for each connection subject to the failure within a specific limit to the corresponding application. In addition, network resources should be managed in such a way that the connections with a high-quality restoration requirement must be allocated with fast and prioritized restoration services, while the connections with lower restoration requirements can be equipped with slower restoration services (which is possibly more capacity efficient). Therefore, the development of the network resiliency schemes that can guarantee the restorability and also address spare capacity allocation efficiency and the maximum allowable interruption time should be the most critical issue in this NGN project. That is also the main reason that Canterbury Team highlights it several times and is continuing work on it. Before the Shared Segments Protection (SSP) mechanism is presented, the critical resiliency constraint imposed by the Shared Risk Link Group (SRLG) is firstly described and the conventional shared protection schemes (including shared path- and linkprotection) are also briefly overviewed.

3 1.1 Shared Risk Link Group (SRLG) Constraint Shared Risk Link Group (SRLG) [1] is defined as a group of network elements (i.e., links, nodes, physical devices, software/protocol identities, or a mix) subject to the same risk of single failure. The SRLG constraint defines the availability of protection resources to a working path, which stipulates that any two working paths sharing the same risk of failure (or in the same SRLG) cannot make use of the same protection resources. The SRLG constraint is imposed on the selection of protection resources for a newly arrived working path, which marks some of the existing protection paths as prohibited avoiding a resource conflict during a restoration process after failure. The purpose of following the SRLG constraint is to guarantee 100 percent restorability for failure on any single link or node in the network. An example demonstrating the SRLG constraint is given in Figure 1 below. Figure 1 An illustration of SRLG Since W2 traverses the link A-B, which shares the same risk of single failure with W1, the protection path for W2 should exclude the possibility of using any of the protection resources used by W1. Otherwise, a failure on link A-B will result in a resource conflict between W1 and W2 when both paths switch their traffic to the same protection channel. Therefore, the SRLG constraint stipulates that W2 cannot take any network resources along P1 for protection purposes. It is clear that as W1 becomes longer, there would be more working network components belonging to the same SRLG that suffers the sharing constraint. The SRLG constraint can be relaxed if extra switch-merge node pairs are allocated along a working path (or divide a large SRLG into several small ones) so that the sharing of protection resources can be improved. 1.2 Shared Path Protection For path-based protection, the source node of a working path computes a protection path by ensuring that the protection path is diversely routed from the working path according to the SRLG constraint. If a fault occurs on the working path, the terminating node in its control plane realizes the fault and sends a notification indicator signal (NIS) to the first hop node of the path to activate a switchover. The source then immediately sends a wake-up packet to activate the configuration of the nodes along the protection path and then switches traffic over from the working path to the protection path. The restoration time of path protection is strongly determined by the total length of the

4 working and protection path segments that circumvent the failed network element. Although path-based protection yields a simple signaling mechanism by circumventing any failure in an end-to-end fashion, it cannot guarantee the failure recovery time for the working path that need to meet stringent requirements on service continuity. In addition, with the path-based protection scheme, the SRLG constraint may limit resource sharing without any relaxation, and as a result impair performance. 1.3 Shared Link Protection Link-based protection was originally devised for ring-based network architectures such as synchronous optical network (SONET), where network planning efforts significantly influence performance. In general, link-based protection in mesh networks is defined as a protection mechanism that performs fault localization during the occurrence of a failure, restores the interrupted services by circumventing the traffic from a failed link or node at the upstream neighbor node, and merges the traffic back to the original working path at the downstream neighbor node. With this definition, two switch nodes must be arranged for every node along a working path. A. link-based protection provides the fastest restoration due to fault localization and better throughput due to the relaxation of the SRLG constraint. However, the downstream neighbor node and link are required to have separate protection segments, which may impair performance by consuming extra protection resources. 2. Framework of Shared Segments Protection (SSP) 2.1 Shared Segments Protection The promising protection scheme, i.e., Shared Segments Protection (SSP) [2] is an end-to-end service-guaranteed shared protection scheme, which enhances the link- and path-based shared protection to provide finer service granularity and higher network throughput. The main idea of SSP is to subdivide a working path into several reasonable length and overlapped segments when the working path is allocated. Each of the segments forms a protection domain (or called P domain), which has a local significance to the working path as shown in Figure 2 below: Figure 2 SSP scheme with overlapped P domains

5 The overlap between adjacent protection domains is for the purpose of protecting node failure along a working path. Each P domain has a path switch LSR (PSL) and a path merge LSR (PML), which switches over and merges back the affected traffic during a failure, respectively. For example, the PSL in each of the three P domains in Figure 2 above is A, E, and I, respectively; the PML is F, J, and N, respectively. A disjoint-routed protection path segment is searched for each working path segment in a P domain. Unlike the conventional resilience schemes, SSP performs the restoration process within a predefined P domain instead of along the whole path. With each working path segmented, restoration service can be guaranteed by limiting the size (or the sum of the distance of working and protection path segments) of P domains. Compared with path-based protection, the segmentation of working paths also yields less computation latency during path selection by using a fully distributed computing process. With this, the task of end-to-end diverse routing is divided into several subtasks, each of which deals with much reduced amount of link-state and computation efforts in the PSLs of the working path. The advantages of the SSP framework over the ordinary protection schemes (path-based and link-based shared protection) are summarized as below: Guaranteed recovery time: Both the notification and the traveling of the wake-up message are performed within a P domain; therefore, the restoration time can be guaranteed by adjusting the size of the P domains along a working path. Scalable: The computational complexity of protection paths is simplified due to the segmentation of working paths. The protection domain allocation process is scalable to the length of working paths (and also to the network size). Capacity Efficient: Compared with link-based shared protection, SSP provides flexibility and capacity efficiency to yield protection services adaptive to the requirements. SRLG Effective: Compared with the end-to-end path-based shared protection, the total computation complexity for correlating the SRLG constraint can be reduced due to the segmentation of working paths. In addition, more resource sharing can be explored by relaxing the SRLG constraint. 2.2 SSP Recovery Time Model The SSP recovery time model is extensively studied in [3]. The average recovery time [ms] expressed in ms (i.e., ) for SSP just in WDM layer can be modeled as follows: t r

6 t [ ms] r = f + 10n + n 1 j= l [ km] i + 10 P k + P k i= l where f is the number of lightpaths interrupted by one link failure. Variable n denotes the number of OXC nodes, from (and including) the upstream node adjacent to the failure, to the branch node of the recovery segment. The length of the ith link of the segment [km] (measured in km), after the branch node and before the merge node, is denoted by l i. The recovery time for shared path protection can also be represented using this analytical formula by setting the branch node (respectively the merge node) as the source (respectively destination) node. The term P is the number of links on the k-th protecting segment of the failed k-th working segment. More detailed information on SSP modeling and the methods of how to calculate its spare capacity allocation and recovery time could be found in [2], [3]. The numerical results significantly show that the restoration time of SSP can be shortened and guaranteed. In addition, a higher possibility of resource sharing can happen between different protection segments under considering the SRLG constraint. In summary, we want to highlight that, in our understanding, the main purpose of this NGN resiliency project is to develop a systematic methodology to quantitatively estimate a connection's availability, especially when various dedicated or shared protection and restoration schemes are applied to the connection. Such a methodology can essentially help Telecom to understand how well and at what cost a connection is protected and whether or not a service quality can be guaranteed instead of simply stating that a connection is protected. As a consequence, the main outcome of this NGN project should be to determine cost-effective, service availability-aware and connection provisioning schemes for differentiated services. Shared Segments Protection (SSP) is a promising solution to achieve this goal by flexibly adjusting segment sizes. It needs to be further studied in the light of the previous proposal of using the outage/recovery time distribution in the resiliency model, which is an applicable model for holistically evaluating several resilience issues, i.e., recovery time, spare capacity allocation and service availability. Discussion Points Point 1 The Canterbury team suggests further exploration of the shared segment protection scheme in this NGN project, for possible consideration as one of selected resilience mechanisms for the NZ Telecom NGN. Point 2 The Canterbury team also suggests verification of the SSP recovery time model in the NGN measurement plan. k [ km] i

7 References [1] D. Papadimitriou and et al., Inference of shared risk link groups, in Internet draft, Nov [2] P.-H. Ho, J. Tapolcai, and T. Cinkler, Segment shared protection in mesh communication networks with bandwidth guaranteed tunnels, IEEE/ACM Trans, VOL.12, NO.6, December, 2004 [3] J. Tapolcai, P.-H. Ho, D. Verchere, and T. Cinkler, A Novel Shared Segment Protection Method for Guaranteed Recovery Time, in Proc. BROADNETS 2005, Boston, MA, Oct. 3-7, 2005