Fault-tolerant design of packet switched network with unreliable links

Abstract

Network optimization and design procedures often separate quality of service (QOS) performance measures from reliability issues. This paper considers channel allocation and flow assignment (routing) in a network subject to link failures. Fault-tolerant channel allocation and flow assignments are determined which minimize network cost while maintaining QOS performance requirements. This approach is shown to yield significant network cost reductions compared to previous heuristic methods.

Introduction

The joint capacity and flow assignment optimization problem plays a central role in the design of large scale packet switching networks 1, 2, 3, 4. This design methodology determines the capacity allocation and flow or routing assignment to optimize network cost subject to quality of service (QOS) performance constraints such as average delay, throughput or buffer overflow probability. In conventional capacity and flow assignment optimization problems, network reliability issues are seldom addressed. The common approach in network optimization procedures is to separate QOS requirement aspects from reliability issues, or vice versa. Network reliability clearly affects a quality of service or performance objective directly and thus, reliability issues should be integrated with other QOS measures in the course of network planning and design. In the tactical battlefield environment, for example, fault-tolerant network design methodologies are of particular importance. It is also important to develop fault-tolerant schemes for large-scale high-speed networks serving critical applications.

The concept of performance-related reliability measures was considered in [5] for analyzing a degradable computing system. A joint measure of performance and reliability has been formulated into a performability measure, which is defined to be the steady-state probability that the network is in a set of states, in which the performance measure is within a specified range [5]. Using the performability measure, [6] proposed a minimum-cost dimensioning approach for common channel signaling networks under both performance and reliability constraints. A network reliability measure based on a routing model was proposed for circuit switched networks in [7] where reliability is defined to be the difference between the amount of lost call traffic under no link failure and that in the presence of link failure. A fault-tolerant joint capacity and flow assignment design approach, called proofing method, has been proposed in [8]. This approach assigns redundant capacity, beyond that required to maintain a QOS level under normal network conditions, to each link in advance. This capacity augmentation prevents performance degradation in the presence of network state changes caused by link failures. Capacity augmentation approaches in the proofing method were based on heuristic considerations and were not claimed to be optimum in any sense. Moreover, the proofing method itself does not reflect the cost optimization issue.

In this paper, we generalize the proofing method by formulating an optimal capacity augmentation and flow assignment (OCA/FA) problem. The objective is to obtain fault-tolerant capacity and flow assignment schemes in which traffic demand and QOS network performance requirement are met at a minimum network cost regardless of the network topological state. This OCA/FA problem is generally a nonlinear convex optimization problem. A novel integer solution technique based on marginal analysis is devised to obtain a suboptimal capacity augmentation scheme while the optimal flow assignments in each state are determined using the optimal routing algorithm. Comparisons to previously proposed heuristic capacity augmentation procedures in [8] show that a significant reduction in network cost can be obtained by using this optimization approach.

This paper is organized as follows. The network design model with unreliable links and the QOS performance measures are introduced in Section 2. The fault-tolerant OCA/FA problem is formulated in Section 2. The proposed solution technique is given in Section 3while numerical examples are given in Section 4.