Reliable Topology Design of Wireless Networks under Correlated Failures

Abstract

Inherent vulnerability of wireless backhauling to random fluctuations of the wireless channel complicates the design of reliable backhaul networks. In the presence of such disturbances, network reliability can be improved by providing redundant paths between given source and destination. Many studies deal with modifying and designing the network topology to meet the reliability requirements in a cost-efficient manner. However, these studies ignore the correlation among link failures, such as those caused by rain. Consequently, the resulting topology design solutions may fail to satisfy the network reliability requirements under correlated failure scenarios.To address this issue, this paper studies the design of reliable wireless backhaul networks under correlated failures with focus on rain fading. We consider green-field topology design and brown-field topology upgrade scenarios with the objective to minimize the total cost of wireless links added to meet the target reliability requirement in the presence of correlated link failures. We propose a new model to formulate the spatial correlation using pairwise joint probability distribution of rain attenuation between different links. This model is applied to consider the linkwise correlation along individual paths, as well as the correlation among the multiple redundant paths from the source to the destination node of a traffic flow. We formulate the problem as a quadratic integer program, which is NP-hard, and develop a heuristic algorithm to find near-optimal solutions. Performance evaluation shows that correlation-aware design improves the resiliency under rain disturbance at a slightly increased cost.

A Realistic Open-Data-based Cost Model for Wireless Backhaul Networks in Rural Areas

“…in [12] an optimal network planning and cost assessment tool is developed for 5G networks. In [13] an optimization model is used to formulate a backhaul design maximizing reliability for a channel model that includes rain attenuation. However, 5G has a specific focus on increasing the user performance in dense areas, but no specific provision for under served rural areas [14].…”

“…Ideally, we could formulate an optimization problem that contains all the constraints and find an optimal solution. However, this approach can scale up to tens of nodes (like in previous works [13]) while in our setting we have thousands of nodes and tens of thousands of potential edges. The goal of our paper is not to simplify the problem to make it tractable with some theoretical formulation that allows to reach some global optimum, or to introduce new approximation algorithms to marginally improve the algorithmic state of the art, but to provide a readily usable solution for network planning and engineering in arbitrarily large graphs, to study the feasibility of the mixed mesh/wired approach.…”

WIP: Analysis of Feasible Topologies for Backhaul Mesh Networks

“…The third metric is a state of the art algorithm for topology design [11]. This work tackles the problem of designing a robust backhaul topology using a realistic model for the probability of link availability under heavy rain or snow.…”

“…We focus on robustness metrics of the graphs realized with TrueNets and we compare them with the properties of a full-mesh graph realized on the same set of nodes. The motivation comes from recent works that deal with mesh backhaul assuming a full mesh between the network nodes [1], [11] and we want to test how this assumption influences the results in a more generic setting.…”