A Fluid Model for Mobile Data Offloading Based on Device-to-Device Communications with Time Constraints

Proximity-based content sharing between nearby devices in cellular networks using device-to-device (D2D) communications—without routing through the base station—has emerged as a promising solution for offloading traffic from the core cellular network and reducing network congestion, especially when the users requesting content can tolerate some delay before receiving it. Although several analytical models have been developed to derive theoretical performance bounds of D2D-based offloading schemes under different user mobility patterns and routing algorithms used for content dissemination, how to jointly analyse time-limited caching and forwarding policies with both constant and asynchronous timeouts remains still an unsolved problem. To address this issue, we propose a novel fluid model based on ordinary differential equations (ODEs) for the performance analysis of a general D2D-based mobile data offloading scheme, called OORS, which considers both content delivery guarantees and time limitations for storing content copies in local device caches, making it more practical for real-world applications. We also formulate an optimisation problem to maximise the utility of the content dissemination process through a simplified analysis of the stationary regime of the ODE model. Simulation results validate the accuracy of our model predictions, in terms of both aggregate statistics and the temporal evolution of the system state, using both synthetic and real-world mobility datasets. Finally, we compare OORS—optimally tuned with respect to protocol parameters—to two state-of-the-art content offloading schemes, Push-and-track (PAT) and SNSNI, a seed node selection algorithm based on node influence. Our results show that OORS achieves similar offloading efficiency to the benchmarks while reducing the number of content copies by at least 50%.