Improving the Survivability of a Stochastic Communication Network
Abstract
This research examined the performance of communication networks with stochastically failing components; it also investigated investment strategies to improve network performance. The performance of the network was measured in terms of the amount of information that could by handled. Since the exact calculation of expected maximum flow, which measures the performance of a stochastic network, is mathematically intractable, the bounds of expected maximum flow were investigated. The network performance, both under normal and adverse conditions, was measured analytically by formulating the problem as a linear programming model. The improvement of network performance was made by increasing the capability of the components to handle the information. Mixed integer programming models were developed to determine the investment strategy which maximized the lower bound of expected maximum flow. Because of the complexity involved in developing these models for a large network, a Prolog program was written to generate formulations for all performance and investment strategy models based on the arc-path incidence matrix. Such formulations serve as a direct input file to off-the-shelf mathematical programming packages such as the LP/MIP-83 linear and mixed integer programming system. In this research, three realistic communication networks of various sizes and topologies were analyzed.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1988
- Accession Number
- ADA202872
Entities
People
- Eugene Yim
Organizations
- Air Force Institute of Technology