Characterization of the End-to-End Delay in Heterogeneous Networks

Abstract

The network slicing architecture is proposed to split single or even multiple physical systems into multiple private logical ones and provide a different Quality of Service (QoS) level for various applications based on their requirement. An important parameter of QoS is the end-to-end (E2E) delay. From a network management viewpoint, E2E delay models play a crucial role to decide what kind of technology can support the required quality of service. In this paper, we investigate the E2E delay distribution for different communication networks to obtain a comprehensive method to model it. We compare the distribution of the E2E delay for heterogeneous networks with multiple known distributions. Moreover, we quantify the error of adopting known distributions to represent the data in a unified way. The results obtained for the different datasets show that the Generalized Extreme Value (GEV) distribution is the approximation exhibiting the lowest error, being a candidate to model the E2E delay for a large variety of scenarios.