Load balancing of elastic data traffic in heterogeneous wireless networks

Heterogeneous wireless networks have been introduced to enable cellular networks to handle the continuously increasing amount of mobile data traffic. In heterogeneous networks, the deployment of macrocells is accompanied by the use of low power pico and femtocells (referred to as microcells) in hotspot areas inside the macrocell, which increases the data rate per unit area. In this paper, we study the load balancing problem of elastic data traffic in heterogeneous wireless networks focusing on a single macrocell that is supported by a number of microcells. This results in a parallel queueing system where individual cells are modelled as single server processor sharing queues. Both static and dynamic load balancing schemes are developed to balance the data flows between the macrocell and the microcells so that the mean flow-level delay is minimized. The performance of static policies is evaluated by analytical and numerical methods, whereas for the dynamic policies, we need to resort to simulations. Our numerical experiments indicate that all dynamic policies can significantly improve the flow-level delay performance compared to the optimal static policy. Among the implemented policies, the so-called myopic policy appears to be systematically the best. However, also a much simpler policy with respect to the required state information, the modified join-the-shortest-queue policy, is able to achieve almost the same performance. In addition, we observe that the performance gain of most of the dynamic polices (including the myopic one) is at least approximately insensitive with respect to the flow size distribution.