Survivable design in WDM mesh networks
He, Wensheng (2006) Survivable design in WDM mesh networks. PhD thesis, Iowa State University.
Full text available as:
This dissertation addresses several important survivable design issues in WDM mesh networks. We first consider single link failure scenarios. To achieve both high efficiency in capacity utilization and fast restoration are primary goals of survivable design in optical networks. Shared backup path protection has been shown to be efficient in terms of capacity utilization, due to the sharing of backup capacity. However, sharing of backup capacity also complicates the restoration process, and leads to slow recovery. Ring-type protection in mesh topology, on the other hand, has the advantage of fast restoration. The p-cycle scheme is the most efficient ring-type protection method in terms of capacity utilization. Recently, the concept of pre-cross-connected protection was proposed to increase the recovery speed of shared path protection. We overview these protection methods and discuss their failure recovery processes. The recovery time of these schemes are compared analytically. We formulate integer programming optimization problems for three protection methods in static traffic scenario, considering wavelength continuity constraint. We investigate the effect of network connectivity on the performance of capacity utilization of the methods by experimenting on topologies with different average nodal degrees. Dynamically provisioning connections, i.e. lightpaths are established on demand as connection requests arrive at the network and torn down when connections are terminated is becoming more important in backbone transport network. Survivable design for dynamic traffic using p-cycle technique has the potential to achieve both fast recovery and capacity efficiency. We develop a p-cycle based scheme to deal with dynamic traffic in WDM networks. We use a two-step approach. In first step, we find a set of p-cycles to cover the network and reserve enough capacity in p-cycles. By doing this, we provision the network built-in resources to be two parts: protection resources and resources available for accommodating the working traffic. The design also ensures that the p-cycles are preconfigured. In second step, we route the requests as they randomly arrive one by one. We propose two routing algorithms. Compared to the shared path protection, in which a primary path and backup path is determined for each request as it arrives, the p-cycle based protection considers the protection in the network as a whole in one step. The p-cycle based design has the advantage of fast recovery, less control signaling, less dynamic state information to be maintained. To evaluate the blocking performance of proposed method, we compare it with shared backup path protection by extensive simulations. Although the failure of single component such as a link or a node is the most common failure scenario, it is possible to have multiple links fail simultaneously. We consider double-link failure scenario in which two links can fail simultaneously. We propose a path-based protection method for two-link failures in mesh optical networks. We identify the scenarios where the backup paths can share their wavelengths without violating 100% restoration guarantee (backup multiplexing). We use integer linear programming to optimize the total capacity requirement for both dedicated- and shared-path protection schemes. Numerical results indicate that, backup multiplexing significantly improves the efficiency of total capacity utilization. The recently proposed light trail architecture offers a promising candidate for carrying IP centric traffic over optical networks. The survivable design is a critical part of the integral process of network design and operation. We propose and compare two protection schemes, namely connection based protection and link based protection, that can achieve 100% protection against single link failure. The survivable light trail design problem using connection based protection model is solved using a two-step approach.
Archive Staff Only: edit this record