Reliability computation of an IP/ATM network with congestion

A state space based algorithm was proposed to compute the reliability of IP/ATM networks at IP level including congestion. The algorithm is based on the assumption that packet rerouting in the network is only caused by physical link failures. This assumption makes it possible to calculate packet transmission paths and congestion at IP routers on a state by state basis. Congestion is measured by the probability of buffer overflow at routers, and is computed using Markov Chain controlled on-off source model. If a router has a separate pair of queues for each neighboring node, congestion failure resembles link failure; if a router has only one pair of queues for all traffic that goes through it, then congestion failure is a node failure. Based on the bounds |R_all-R_allⁱ|< (qn)ⁱ<ε (with R_allⁱ the approximated all-terminal reliability with less than i simultaneous trunk failures, q trunk failure rate, n the number of trunks), the number of states to be computed is O(nⁱ), with i=log(ε)/log(qn) and qn≪1. The complexity of the algorithm is O(nⁱ*polynomial(m, n)), in is the number of nodes. This algorithm was used to compute the reliability of vBNS-an IP over ATM network, at IP level. The results are useful for network design and planning.