| Summary: | In this paper, we analyze the throughput of data dissemination at the level of users' interests. We show that users' interests have the ability to drastically improve upon existing throughput scaling's established under the assumption that users show the same preference in any type of data they encounter. More precisely, we consider the scenario where each data source estimates the recipients that will be interested in its data based on user interest probability, which is described by a Zipf-distributed data popularity that decays of exponent α with data ranking. For such a user-centric model, we divide our analysis into different cases depending on data catalogue size K and study their respective throughput performance. With totally n users assumed, we present closed-form expressions of user-centric throughput versus n, α, and K. In particular, our results reveal that when α = 1 where users' interests exhibit a moderate level of heterogeneity, the maximum throughput of Θ(√n) (except for a poly-logarithmic factor) can be achieved in all the situations, with appropriate choice of K. The results augment the existing scaling laws derived in network-centric situation, in that given the same throughput data can be disseminated efficiently to more recipients in a user-centric network.
|
|---|
