FAULT TOLERANT PACKET-SWITCHED NETWORK DESIGN AND ITS SENSITIVITY

Reliability and performance for telecommunication networks have been traditionally investigated separately in spite of their close relation. A design method integrating them for a reliable packet switched network, called a proofing method, was discussed in previous papers. This paper first presents the proofing method in detail. Then two heuristic design approaches (max-average, max-delay-link) for optimizing network cost in the proofing method are described. In order to verify their effectiveness and applicability, they are compared numerically for three example network topologies. Finally, the sensitivity of these two methods is examined with respect to changes in traffic demand and in link reliability. Numerical results show that the max-delay-link method provides a lower minimum network-cost than does the max-average method, for both a small and a large example network. The answers obtained by these two methods are not highly sensitive to changes either in traffic demand or in link reliability. Thus a network designed by these two methods is robust to a system change which is not considered at a design stage. The max-average method is superior to the max-delay-link method in terms of the sensitivity. Many other sources of failures must be considered in our failure model, eg, node and software faults. Only statistically independent failures are considered. Statistical dependence effects must be included to make the model more realistic. The max-average and max-delay-link methods cannot prevent a state where there are no routes between a particular source and destination node pair. To cope with this, a topological design method must be added to these methods. The computational complexity of these methods needs to be clarified to identify how large a practical problem can be solved using them.