Packet jitter measurement in communication networks: a sensitivity analysis

Communication networks are widespread today. New services are now starting to run on these networks to meet the requests of modern users. To offer good service typically hard constraints have to be satisfied, especially when real-time services are involved. To this aim the reliable estimation of Quality of Service (QoS) parameters as bandwidth, packet jitter, latency, one way delay, to cite a few, is a key issue for communication networks setup, monitoring and tuning Generally, the measurement of QoS parameters is a complex process that involves several troubles. Some of them depend on the measurement network features: (i) the medium employed to convey the data stream (optic fiber, wireless, wired, etc.), (ii) the type and number of network devices involved (network interface cards, switches, routers, etc.), (iii) the type of communication and access protocols, and (iv) the variability of the network background traffic which could strongly influence the measurement result. Further troubles come from the measurement chain and the measurement method adopted: type of measurement instruments, sampling and observation times, packet size, type of measurement post processing to cite a few. Therefore, they can be thought as sources of measurement uncertainty able to strongly influence the measurement process and the reliability of the results. With reference to the measurement chain and the measurement method adopted, in this paper a deep sensitivity analysis aimed to identify the main quantities influencing the measurement results and to study their modeling as uncertainty components is performed. The attention has been focused on the packet jitter estimation made on a real test bed: the fiber optic based local area network of the University of Cassino