Performance Analysis of Opportunistic Content Distribution via Data-Driven Mobility Modeling

An opportunistic network is formed by co-located mobile users in order to exchange data via direct wireless links when their devices are within transmission range, without relying on the use of fixed network infrastructure. In this thesis we investigate the capabilities of opportunistic networks and cover two main areas: data-driven modeling of user mobility and analytic performance evaluation of location-aware opportunistic content distribution. The first part of the thesis focuses on mobility modeling. We collect a dataset of user associations in the wireless network of the KTH Royal Institute of Technology, and characterize the mobility of users in this dataset both from the network and from the user perspective. From the network perspective, we model the aggregate mobility and access patterns to different parts of the network. To characterize individual mobility, we assess how mobile the users are, and how accurately their movements can be predicted in the near future. Based on these findings, and on the analysis of several other mobility traces, we propose a mobility model for populations with churn, that is specifically tailored for the evaluation of opportunistic content distribution. In the second part of the thesis, we evaluate the performance of opportunistic content distribution in ephemeral, location-aware networks where content is stored only on the user devices within the locale of interest. We develop a framework that allows modeling of the spread of information as a stochastic process and accurate capturing of the stochastic fluctuations in the number of distributed content items. We study the feasibility of opportunistic content distribution and, by means of stochastic stability analysis, assess how the system parameters can be engineered to ensure content persistence. We show that the content persistence strongly depends on the density of users, and that the requirements for user resources are relatively low already for moderate densities.QC 20181016