An approach to the identification of network elements composing heterogeneous end-to-end paths

Today's networks are becoming increasingly complex and the ability to effectively and efficiently operate and manage them is ever more challenging. Ways to provide end-to-end Quality of Service have to cope with the increasing heterogeneity of these networks, which is due to the several actors of current network scenarios, from access networks to end-user devices, from protocols to applications and operating systems. Exploiting such heterogeneity, in this paper we present an approach to the identification of each element composing an end-to-end path. Such identification is useful in several situations. For instance, it can improve the performance of adaptive and network-aware applications, it can help intelligent routing approaches, and it can be used in network and service overlay scenarios. Our approach, based on Bayesian classifiers, utilizes the measurements and off-line processing of three QoS parameters, that are delay, jitter, and packet loss. To illustrate the capabilities of our proposal, we present the results of a large set of experimentations performed with both different sets of features and different sets of QoS parameters. Using measures related to various time periods, in which the considered paths presented diverse characteristics, we show that a blind identification of network bricks is possible and that its results present a good degree of generalizability.