Network Dimensioning with Minimum Unfairness Cost for the Efficiency

摘要

Network dimensioning is a specific kind of optimization problems. In general the main goal in this task is to ensure a connection between given pairs of nodes (source-target) with possible high efficiency. When each pair (demand) brings different revenue, the problem of blocking less attractive demands appears. Usually this situation is caused by not including any fairness criterion into optimization and thus optimizing only the total (revenue) efficiency of the system. Another complication is the fact of inverse proportionality of these criteria. In this paper an optimization model has been examined which takes into account a fairness criterion and minimizes the loss of system efficiency. It may also be understood as optimizing the ratio of fairness degree to the mean of the traffic flow in the network. For implementation of the model the CPLEX package was used. As input data the example of backbone Polish network structure was chosen. To evaluate the approach, basic statistics which help in describing the equity of distribution such as standard deviation, kurtosis and Gini coefficient are used.