Capacity Estimation for Self-Backhaul in mmWave Networks

We consider the performance implications of using self-backhaul in a millimeter wave (mmWave) wireless network. In a self-backhauled network some nonanchored basestations route their traffic through another anchored basestation rather than utilize an expensive wired backhaul. Self-backhaul is particularly attractive in the mmWave context since we can form narrow beams that reduce the interference due to the backhaul links. In earlier work, Singh et al. demonstrated that we can maintain a fixed level of performance while reducing the density of wired backhaul as long as we simultaneously increase the total density of basestations. Singh et al. assumes a static configuration in which scheduling gains are not exploited. In this work we study how interference and load affect the overall performance of self-backhauled networks and we examine which network parameters determine whether self-backhaul is beneficial. We demonstrate that highly directional beams and a steep path loss function improve the tradeoffs associated with self-backhaul. We also develop a scaling law that allows us to easily estimate the self-backhaul capacity. We conclude by presenting a scheduling algorithm for selfbackhauled networks and evaluating the scheduling gains that are feasible. A key difficulty with the scheduling is that we must respect the half-duplex constraints under which a non-anchored basestation cannot receive and transmit at the same time.