| The technology of communication networks has changed the world. The dependency between human life and communications is becoming closer and more complicated. The requirements of hardware and software to support Quality of Service (QoS) for communications are more critical. The available bandwidth and end-to-end delay are two major factors affecting QoS. To satisfy these requirements, high data rate networks (e.g., OC-192 with 10Gbps data rate) are deployed widely now. Even though these high-speed networks can satisfy the heavy traffic requirements, there is an essential issue associated with the backbone trunk to ensure the QoS during the network failures. Because each single cable carries a huge amount of data, it is a serious problem if a link is broken. It usually takes days or weeks to repair a backbone link breakage. During this period, affected traffics have to be rerouted through alternate paths. Otherwise, the QoS suffers from the lack of network resources. The survivability of the network is the capability to restore the affected transmission when network failures occur. A good design for the survivable network can effectively reroute traffic to avoid interruption, with the minimum cost.; Label switching networks play the major role in today's backbone networks. Two examples of label switching networks are the ATM network and the emerging MPLS network. We focus on these two techniques to show how to implement the path restoration strategy to achieve effective control, fast restoration, and economic network capacity placement. Two kinds of design problems are addressed in this dissertation: the predetermined working flow problem and the joint-optimization problem. The mathematical models for the survivable networks are discussed in the forms of Linear Programming (LP) and Mix Integer Programming (MIP). We proposed a compound algorithm containing the rounding method, the MIP method, and the heuristic method to solve the problems in different ways. The refining process is introduced to utilize the residual capacity for the expandable demand. This technique effectively reduces the wasted capacity in the networks and further creates benefit without spending extra capital. An overall algorithm is proposed for the final decision making. Numerical examples are given to illustrate the procedure and performance. |
