5G access-link provisioning and coordination: Tradeoff between proactive and reactive strategies

We explore the effect of imperfect knowledge on multipoint downlink traffic engineering (TE) (with dynamic point selection DPS) computed centrally and jointly with power control (PC) optimization versus distributed PC techniques. Instead of considering a single metric that shows qualities in a single dimension, we observe both performance and resilience by observing how many users get quality of experience (QoE) — or otherwise, how many users would complain that they do not receive the expected service. We identify that centralized traffic engineering provides the best performance (highest class of service) with distributed power control in a practical (imperfect) network. Although this has a clear drawback as it comes with a low resilience in face of bad service admission. We conclude that fifth generation networks will need to be able to dynamically change optimizing strategies to face different loading of the networks.

I. Introduction

New virtualized technologies are considered for future fifth generation mobile networks. In the mobile wireless access link layer, the notion of cloud radio access network [1] (CRAN) is considered. In effect, a set of radio access points (AP) are virtualized and seen as one cell associated to a user-equipment (UE). In this study we consider a simpler version for simulation purposes, with multiple cells serving one same UE dynamic point selection (DPS), although transmission is not coordinated in between each.