An efficient experimental approach for the uncertainty estimation of QoS parameters in communication networks

In communication networks setup and tuning activities, a key issue is to assess the impact of a new service running on the network on the overall Quality of Service. To this aim suitable figures of merit and test beds have to be adopted and time-consuming measurement campaigns generally should be carried out. A preliminary issue to be accomplished for is the metrological characterization of the test set-up aimed to provide a confidence level and a variability interval to the measurement results. This allows identifying and evaluating the intrinsic uncertainty to be considered in the experimental measurement of Quality of Service parameters. This paper proposes an original experimental approach suitable for the purpose. The uncertainty components involved in the measurement process are identified and experimentally quantified by means of effective statistical analyses. The proposed approach takes into account the general characteristics of the network topology, the number and type of devices involved, the characteristics of the current services operating on the network, and of the new services to be implemented, as well as the intrinsic uncertainties related to the set-up and to the measurement method. As an application example, the proposed approach has been adopted to the measurement of the packet jitter on a test bed involving a real computer network displaced on several kilometers. The obtained results show the effectiveness of the proposal.