Modeling and analyzing device-to-device content distribution in cellular networks
MetadataShow full item record
Device-to-device (D2D) communication is a promising approach to optimize the utilization of air interface resources in 5G networks, since it allows decentralized proximity-based communication. To obtain caching gains through D2D, mobile nodes must possess content that other mobiles want. Thus, devising intelligent cache placement techniques are essential for D2D. The goal of this dissertation is to provide randomized spatial models for content distribution in cellular networks by capturing the locality of the content, and additionally, to provide dynamic content placement algorithms exploiting the node configurations. First, a randomized content caching scheme for D2D networks in the cellular context is proposed. Modeling the locations of the devices as a homogeneous Poisson Point Process (PPP), the probability of successful content delivery in the presence of interference and noise is derived. With some idealized modeling aspects, i.e., given that (i) only a fraction of users to be randomly scheduled at a given time, and (ii) the request distribution does not change over time, it has been shown that the performance of caching can be optimized by smoothing out the request distribution, where the smoothness of the caching distribution is mainly determined by the path loss exponent, and holds under Rayleigh, Ricean and Nakagami fading models. Second, to take the randomized caching model a step further, a spatially correlated content caching scenario is contemplated. Inspired by the Matérn hard-core point process of type II, which is a first-order pairwise interaction model, D2D nodes caching the same file are never closer to each other than the exclusion radius. The exclusion radius plays the role of a substitute for caching probability. The optimal exclusion radii that maximize the hit probability can be determined by using the request distribution and cache memory size. Unlike independent content placement, which is oblivious to the geographic locations of the nodes, the new strategy can be effective for proximity-based communication even when the cache size is small. Third, an auction-aided Matérn carrier sense multiple access (CSMA) policy that considers the joint analysis of scheduling and caching is studied. The auction scheme is distributed. Given a cache configuration, i.e., the set of cached files in each user at a given snapshot, each D2D receiver determines the value of its request, by bidding on the set of potential transmitters in its communication range. The values of the receiver bids are reported to the potential transmitter, which computes the cumulated sum of these variables taken on all users in its cell. The potential transmitter then reports the value of the bid sum to other potential transmitters in its contention range. Given the accumulated bids of all potential transmitters, the contention range and the medium access probability, a fraction of the potential transmitters are jointly scheduled, determined by the auction policy, in order to optimize the throughput. Later, a Gibbs sampling-based cache update strategy is proposed to iteratively optimize the hit rate by taking the scheduling scheme into account. In this dissertation, a variety of distributed algorithms for D2D content caching are proposed. Our results indicate that the geographic locality and the network parameters have a significant role in determining and optimizing the placement strategy. Exploiting the user interactions and spatial diversity, and incentivizing cooperation among D2D nodes are crucial in realizing the full potential of caching. Furthermore, from a network point of view, the scheduling and the caching phases are closely linked to each other. Hence, understanding the interaction between these two phases helps develop novel dynamic caching strategies capturing the temporal and spatial locality of the demand.