2. Consideration for PON protection planning

#### 2.1. Network topology

in terms of reach, bandwidth, and the number of subscribers. In PON, all services are originated from an optical line terminal (OLT) at the central office (CO). End-face of the OLT is connected to a 15–20 km feeder fiber (FF) that extends the network toward the subscriber premises called optical distribution network (ODN). Remote node (RN) receives the FF at ODN, which houses a 1 : N bidirectional passive optical coupler (POC). N output ports from the POC are fed into short-branched distribution fibers (DFs) that connect the RN to individual

PON has emerged as a promising candidate to resolve the last-mile bottle, owing to its signifi-

• Support for high network capacity in terms of reach, bandwidth, and the number of subscribers due to a complete optical fiber (OF) path between OLT and ONU modules. • Minimum capital expenditure (CAPEX) by sharing FF between OLT and multiple ONUs. • Reduced operational expenditure (OPEX) through passive components at RN, which

• High degree of flexibility, owing to the use of FF between OLT and multiple subscribers. With the rapid increase in PONs capacity, fault detection and restoration at satisfactory costs have turned the network reliability to a new challenge for Internet service providers. Each subscriber is interested in seamless reception of maximum bandwidth at minimum possible cost. However, the conventional PON architecture has limited protection, which results in significant data loss at the event of failure in optical components including OF medium. Therefore, it is imperative to devise an architecture, which is capable of maintaining a seamless flow of upstream and downstream

Two techniques are readily adapted to provide fault detection and restoration in PON, namely pre-planned and dynamic protection. The latter relies on fault detection and restoration through diagnosis at the higher levels and dynamically allocates resources at the event of failure. Such technique requires more time for traffic restoration between OLT and ONU modules, as upper layer recovery techniques usually utilize routing tables, topology recalculations, and slow convergence time. Yet there is no guarantee for fault restoration at the physical layer [1–3]. Therefore, for the facilitation of an effective and prompt fault detection and restoration, it is highly desirable

Pre-planned protection utilize an optical-layer approach by providing dedicated backup paths for components including OF medium. This type of protection is planned at the network design phase, owing to the fact that topology of PON remains same, and the proposed solution can address fault restoration at both feeder and ODN. This type of protection provides high reliability at minimum recovery time in the event of failures at both optical components and OF medium. However, path and resources duplication significantly elevates the CAPEX at the network deployment phase [4, 5]. Therefore, it is imperative to encompass the following consid-

traffic at required capacity and acceptable costs for a common end subscriber [1, 2].

optical network unit (ONU) transceiver modules [1].

requires no power, minimum maintenance and planning.

• Highly scalable as new subscribers can easily join the network.

• Smooth service upgradability with existing infrastructures.

to provide protection measures at the optical layer.

erations while designing a pre-planned protection architecture.

cant advantages like:

340 Optical Fiber and Wireless Communications

Network topology significantly effects the design, redundancy, and deployment cost for the PON. Two common network topologies are used for the deployment of PON, namely tree and ring [1]. In tree topology, the optical signal sent from OLT is divided into N equal parts and delivered to designated ONU modules through respective DFs. Such deployment can provide the required bandwidth at desired number of subscribers; however, a single cut or failure at the feeder level can cripple the entire network by disconnecting the working OLT from ONU modules. Moreover, failure at the DF can also result in significant data loss and high customer dissatisfaction. Therefore, such topology requires the provision of redundancy at both levels of PON, which is achieved by duplicating the network components.

Ring topology is adapted to minimize the cost incurred by the provision of redundant paths in the conventional PON. It utilizes a single ring-based fiber that connects the OLT directly to all ONU transceiver modules. This significantly reduces the effect of fiber cuts or failures [6, 7]. Ring topology provides the required reliability at acceptable costs; however, use of the POC between OLT and individual ONU module introduces serious power budget issues, which effects network capacity in terms of the number of subscribers [4, 8, 9]. Besides the commonly used ring- and tree-based network architectures, hybrid topologies are readily adapted for the implementation of survivable PON at the access domain. These architectures utilize a combination of tree- and ring-based architectures with subsequent topologies such as tree-ring, treestar, ring-star, and bus. Hybrid architectures have proved as a promising candidate to provide the required redundancy at desirable network capacity [10–14].

#### 2.2. Resources to be protected

A typical PON primarily comprises two types of resources that require protection for efficient delivery of information between OLT and ONU modules, namely OF medium and optical components. Both significantly effect the flow of upstream and downstream traffic throughout the network. Figure 1 shows connection availability for PON components based on Table 1 [15]. It is observed that active and passive devices, such as OLT, ONU, POC, X : NPOC, and so on, provide desirable (5 nines) connection availability over the network lifetime, since the rate of failure for these components is significantly low. Furthermore, the mean time to repair (MTTR) for the in-house optical components is minuscule as compared to the on-field components like OF medium that constitutes a major portion of PON architecture and is more prone to failures as shown in Figure 1.

Therefore, OF paths require more attention as compared to other components of the networks, in order to ensure seamless transmission of information, minimize the loss of data, service interruption penalty cost, and PON downtime per year [4].

#### 2.3. Number of subscribers

Number of subscribers refer to the amount of users that a PON can accommodate without compromising the reach and provision of nominal bandwidth. It is an important parameter

Figure 1. Connection availability of basic optical components in PON.


Table 1. Components description, connection availability, and cost.

since it is directly associated with the extent and cost of the network. Number of subscribers is primarily effected by the type of topology at both feeder and ODN along with devices at the CO and RN. For example, a typical tree-star topology can accommodate more subscribers as compared to a conventional ring-based architecture due to the use of 1 : N POC. Whereas, the latter utilizes symmetric Y, 1 : 2 or 2 : 2, POC per subscriber, which introduces a power-budget loss of �3 dBm in each symmetric POC, PPOC ¼ 10Log10 1 2 [6, 16]. This significantly effects capacity of the network since nominal received power is required for high bandwidth communication. Therefore, it is imperative to consider these features at the network planning phase, so that the proposed PON can accommodate maximum number of subscribers at desirable capacity and cost.

#### 2.4. Cost and complexity

Deployment/operational costs and feasibility of PONs primarily depend on complexity of the network architecture. For example, some protection mechanics utilize redundant transceivers at both OLT and ONUs, like ITU-T type C and D, in order to avoid 1 : 1 or 1 þ 1 switching [17]. Although such techniques provide an abrupt recovery to maintain a smooth flow of information between OLT and ONU modules, they significantly elevate the deployment cost of the network. Since more CAPEX is spent on OLT duplication as compared to the 1 : 1 or 1 þ 1 switching, a trade-off must be made between the cost and recovery time at the event of failure. Therefore, it is desirable to minimize the overall system complexity, without compromising the fault detection and restoration time.
