Distributed caching in device-to-device networks: A stochastic geometry perspective

Increasing spatio-temporal correlation in the data demand makes it attractive to cache popular content directly on the user devices so that it can be delivered on demand to the neighboring devices through device-to-device (D2D) communications. Due to the limited storage capacity, each device can only cache a part of the popular content resulting in a distributed caching network. With device locations modeled as a Poisson Point Process (PPP), we assume that the file of interest for a typical device is partitioned into several file portions with each device caching one of the portions randomly. Two equivalent viewpoints for the analysis of coverage probability and average delay in a distributed caching network are discussed for varying cache probabilities and device activity factors.