Optimal downlink-uplink user association in HetNets with traffic differentiation

The widespread adoption of tablets and smartphones, and an abundance of data-hungry mobile applications, are overwhelming wireless networks with increased demand and introduce considerable traffic diversity. Operators struggling to continuously add capacity and upgrade their architecture have resorted instead to building denser deployments to improve spectral efficiency. By increasing the number of cells a user can associate with, (i) user quality of service (QoS) can be improved, and (ii) traffic can be offloaded from congested base stations, to achieve better load balancing. However, these two goals are not always aligned. To this end, we develop an analytical framework for optimal user association in future HetNets that investigates the potential tradeoffs between user- and network-related performance, in a more realistic setup encompassing additional key features: (i) different types of user flows, and (ii) uplink and downlink performance. We believe this better reflects the diversity of the services offered to users and their impact on

system performance. We evaluate our proposed framework through extensive simulations, and provide some qualitative and quantitative insights on the related tradeoffs.