Self-organized backpressure routing for the wireless mesh backhaul of small cells

Abstract

The ever increasing demand for wireless data services has given a starring role to dense small cell (SC) deployments for mobile networks, as increasing frequency re-use by reducing cell size has historically been the most effective and simple way to increase capacity. Such densification entails challenges at the Transport Network Layer (TNL), which carries packets throughout the network, since hard-wired deployments of small cells prove to be cost-unfeasible and inflexible in some scenarios. The goal of this thesis is, precisely, to provide cost-effective and dynamic solutions for the TNL that drastically improve the performance of dense and semi-planned SC deployments. One approach to decrease costs and augment the dynamicity at the TNL is the creation of a wireless mesh backhaul amongst SCs to carry control and data plane traffic towards/from the core network. Unfortunately, these lowcost SC deployments preclude the use of current TNL routing approaches such as Multiprotocol Label Switching Traffic Profile (MPLS-TP), which was originally designed for hard-wired SC deployments. In particular, one of the main problems is that these schemes are unable to provide an even network resource consumption, which in wireless environments can lead to a substantial degradation of key network performance metrics for Mobile Network Operators. The equivalent of distributing load across resources in SC deployments is making better use of available paths, and so exploiting the capacity offered by the wireless mesh backhaul formed amongst SCs. To tackle such uneven consumption of network resources, this thesis presents the design, implementation, and extensive evaluation of a self-organized backpressure routing protocol explicitly designed for the wireless mesh backhaul formed amongst the wireless links of SCs. Whilst backpressure routing in theory promises throughput optimality, its implementation complexity introduces several concerns, such as scalability, large end-to-end latencies, and centralization of all the network state. To address these issues, we present a throughput suboptimal yet scalable, decentralized, low-overhead, and low-complexity backpressure routing scheme. More specifically, the contributions in this thesis can be summarized as follows: We formulate the routing problem for the wireless mesh backhaul from a stochastic network optimization perspective, and solve the network optimization problem using the Lyapunov-driftplus-penalty method. The Lyapunov drift refers to the difference of queue backlogs in the network between different time instants, whereas the penalty refers to the routing cost incurred by some network utility parameter to optimize. In our case, this parameter is based on minimizing the length of the path taken by packets to reach their intended destination. Rather than building routing tables, we leverage geolocation information as a key component to complement the minimization of the Lyapunov drift in a decentralized way. In fact, we observed that the combination of both components helps to mitigate backpressure limitations (e.g., scalability,centralization, and large end-to-end latencies). The drift-plus-penalty method uses a tunable optimization parameter that weight the relative importance of queue drift and routing cost. We find evidence that, in fact, this optimization parameter impacts the overall network performance. In light of this observation, we propose a self-organized controller based on locally available information and in the current packet being routed to tune such an optimization parameter under dynamic traffic demands. Thus, the goal of this heuristically built controller is to maintain the best trade-off between the Lyapunov drift and the penalty function to take into account the dynamic nature of semi-planned SC deployments. We propose low complexity heuristics to address problems that appear under different wireless mesh backhaul scenarios and conditions...