Designing telecommunications networks for the reseller market

This paper presents an algorithm for minimizing the monthly cost of telecommunications networks characterized by multichannel queues with forced balking and hierarchical routing plans. The problem is of major interest to resellers of long distance services and corporate network managers who lease their lines from AT&T and Other Common Carriers. The algorithm can be used to select the most economical combination of lines to meet a desired grade of service, set here to an average blockage rate of 1% during the busy hour. Because of the nonlinearities involved, the problem is formulated as a dynamic program and solved with forward recursion. The Erlang loss formula is used to determine blockage in the case of Poisson arrivals, while Wilkinson’s equivalent random theory is adopted when the input traffic is peaked. Model validation is achieved with a SLAM II simulation. The algorithm is applied to the network of National Telecommunications of Austin, which consists of 5 hierarchical levels and 52 candidate nodes, and carries about 2.3 million minutes of monthly traffic. Current services range from FX connections to Specialized WATS. The results indicate that cost reductions of over 18%, or $30000 per month, can be achieved by periodically adjusting trunk group sizes to match the forecast demand.