Collaborative Caching for D2D Content Sharing in 5G

A Master of Science thesis in Computer Engineering by Ansam Elfadil Kamel Abdelsalam entitled, “Collaborative Caching for D2D Content Sharing in 5G”, submitted in May 2021. Thesis advisor is Dr. Rana E. Ahmed. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).

Abstract

Due to a huge number of mobile devices expected to be connected to 5G wireless networks and their expected demand for high data-rate multimedia services, the core network and backhaul links are expected to carry enormous amount of traffic. Caching the most popular files at the network edge and in user’s devices to support user proximity services in 5G will help to offload traffic in the core network and to increase the cache hit probability. Device-to-Device (D2D) communication in 5G can be utilized to share the cached files between any pair of devices with a minimal involvement from the base station. However, there are many challenges that are needed to be addressed including interference management, mode selection, device discovery, contents placement, popularity index calculation, and the non-cooperative situations in D2D. This thesis attempts to solve most of the above-mentioned problems via collaborative content caching and sharing using D2D communication in 5G networks. The primary objective of the research is to maximize the overall system offloading gain and the cache hit probability in downloading the popular file contents. The proposed system model exploits the social-networking concept, assuming the cell structure in a condensed populated area, such as a university campus or an auditorium. We combine the process of content caching and D2D communication in the WiFi range. In particular, joint resource allocation, mode selection, cache placement and replacement for multiple D2D devices are addressed. Data traffic offloading in three modes of operation, self-offloading, D2D offloading, and Base Station-to-Device (B2D) offloading and mode selection algorithm are implemented. Furthermore, we vary the network parameters and the cache model parameters to assess their impacts on the system performance. The performance of the proposed cache scheme is evaluated through extensive simulations and compared with the popular cache scheme and the baseline performance of the random cache scheme, and it is found that the proposed scheme outperforms the two schemes by 9.4% and 20%, respectively, with respect to cache hit probability. The effects of users’ mobility and disconnection on system performance are also investigated.