Reliability of Single, Double and N2R Ring Network Structures

This paper studies the properties of single, double and N2R ring network structures during link errors. The structure of the network infrastructure must be redesigned in order to fulfil the requirements of services using the Internet in the future; hence, N2R structures have been suggested. N2 R structures are found to be superior regarding network properties compared to the other more traditionally ring structures but the deployment of this type of structure is more complicated. Therefore, this paper suggests a deployment scheme, denote d N2R tube deployment, which is a compromise between the easy deployable dobule ring structure, and the better properties of the N2R structure. The properties of single ring, double ring and N2R structures have never been compared during errors in previous research. Nevertheless, this is an important pa rt of selecting the topology of the network infrastructure for the future. This paper compares single, double, N2R and N2R tube structures when one and two errors are introduced. The resul ts show that even when errors are introduced in the network structures, the N2R structure remain superior. Furthermore, this is also valid with a smaller margin for the N2R tube structures. Keywords—Ring network structures, network planning, network reliability, N2R structures and network infrastructu re. I. I NTRODUCTION Recently, several new services using the Internet have been introduced which set higher requirements to the network infrastructure: Online television/video streaming requi res a significant amount of bandwidth; telerobotics [1] [2] requi re low delays; the increasing use of the Internet for reliable s ervices implies that the reliability of the network infrastru ct re must be increased [3]. There exists several other examples but those mentioned state the main challenges for the new network infrastructure: More bandwidth, lower delays and more reliable networks. The bandwidth challenge can be addressed by replacing existing copper lines with optical fibers, which are upgrade abl simply by replacing terminal equipment. The deployment of fiber optical lines is relatively expensive, mainly because of the expenses for digging ducts. Furthermore, the expected operation time of the network infrastructure is very high, which stresses the importance of choosing a suitable topolo gy. Reducing delays and increasing the reliability of the network infrastructure can be achieved by considering other network topologies than those typically used. This implies that it is necessary to deploy a suitable network structure to support the end-users’ demand for increasing reliabilit y. Therefore, it is crucial to choose a suitable network struct ure, which guarantees the demands of the future. Single rings (SR) are often used in network infrastructure since they make routing and restoration uncomplicated, and still have redundancy in case of failure. Recently, more advanced network structures like the double ring (DR) and the N2R structures [8] have been proposed as replacements for th e SR structure. The DR and N2R structures are more complex regarding deployment and routing, but have better network properties [6], which imply higher capacity and lower delay s. In the United States, an average of 1 cable cut per year on 370km [4] of deployed 1 cable can be expected. Cable cuts are the most frequent failure in a widespread network and the mos t time consuming to repair [4]. The fact that it is impossible t o avoid errors in networks stresses the importance of studyin g how the different network structures behave during errors, in order to ensure that the requirements to the networks are sti ll guaranteed. Two basic error types can occur in the network structures: A node can fail, or a link can fail. Node errors can be caused by equipment or power failure, but usually the recovery time or time to start emergency systems is very low [4], compared to the time it takes to locate and repair a link error. Link errors are usually caused by links being dug over. These types of errors will happen randomly in the network with some probability, depending on the size of the network and the way it is deployed. This paper compares certain properties of the SR, DR and N2R network structures in scenarios where link errors occur. In case of a node error the network is divided into two separate parts since the defe ct node is not connected to the network anymore; hence it is not possible to evaluate the structure as a whole. Furthermo r , a traffic generating element is removed from the network in case of a node error; hence, from the network’s perspective t he overall traffic load in the network is reduced, which reduces the impact of node errors. Therefore node errors are not evaluated in this paper. The comparison of the network structures in scenarios with or without errors can support the selection of the network infrastructure topology of the future. 1Deployed in both ducts and masts. The structure of the paper is as follows: Section II “Network topologies” provides an overview of the network structures and previous research. Section III “Methods” describes the methods used to obtain the results, which are described in Section IV “Results”. Section V “Discussion” discusses and sets the results in perspective to real world applications. Furthermore, research topics are suggested. Section VI “Co nclusion” concludes the paper with a brief summing up of the main findings. II. N ETWORK TOPOLOGIES In the SR structure the nodes are placed in a ring, and neighbouring nodes are interconnected by links. Figure 1 shows a SR network structure, where nodes are denoted Ni and i denotes the node number. All links can be described as interconnectingNi and N(i+1) mod p, wherep is the number of nodes in the structure, and 0 ≤ i ≤ p − 1. The number of nodes,p, is any positive integer larger than 2. All nodes in an SR network are connected to two other nodes; thus the nodes in the structure are of second degree. The SR structure does not scale very well since the distances in the structure increase linearly as the number of nodes increase. 10 11 12 13 14 15 16