Infrastructure and spectrum sharing trade-offs in mobile networks

Abstract

Abstract-We assess the fundamental trade-offs between spectrum and radio access infrastructure sharing. Specifically, we analyze three mobile network sharing scenarios: radio access infrastructure sharing, spectrum sharing, and the combination of the two, which we term full sharing. We perform our analysis employing stochastic geometry models, such as the homogeneous Poisson point process and the Gauss-Poisson process, which allow us to vary the spatial correlation in base station placements of different mobile operators. Our major contributions can be summarized as follows. We derive analytical expressions for the coverage probability and the average user data rate for a shared network created as a union of independently distributed networks for each of the three sharing scenarios. For the general base station placement model, we show that infrastructure and spectrum sharing cannot be simply substituted for each other, as there exists a trade-off in the coverage and data rate performance between the two. Moreover, the combination of the two approaches does not simply produce linearly scaling gains, as the increase in the data rate is traded for a minor reduction in the coverage (when compared to infrastructure sharing performed in isolation). We show that the spatial distributions of the networks of the sharing operators have a significant impact on the performance of sharing. We show that respective density of the networks of the two operators influences how the two operators perceive sharing gains. In particular, it is often the case that a larger operator has more to gain from sharing due to disproportionally lower interference suffered from the smaller operator.

Modelling Multi-Operator Base Station Deployment Patterns in Cellular Networks

“…The spectrum sharing scenario provides significantly worse coverage, which should come as no surprise as we assumed worst-case interference. Interestingly, in the case of spectrum sharing we observe a crossover point, which we have reported also in [14], which is related to the amount of clustering in the evaluated patterns. Clearly, a real multi-network provides higher clustering than any of our models can produce.…”

“…Our initial study, see [14], analyses the tradeoff between infrastructure and spectrum sharing. However, one may extend our work also in other ways.…”

“…For analytical tractability, much of the existing work modelling radio access network deployments (e.g., [11], [12], [13], [14]) relies on the homogeneous PPP model, which assumes no correlation in the analysed patterns. The usage of the PPP model can be analytically justified under sufficiently high log-normal shadowing, see [15].…”

“…Having evaluated the statistical match between our proposed point process models and real multi-network deployments, we now turn our attention to wireless network-relevant evaluation based on the radio access network sharing scenarios of: infrastructure, spectrum and full network sharing, as studied in [14]. As a metric for our evaluation we use the coverage probability [26], i.e., the complementary cumulative distribution function (CCDF) of SINR, which we obtain for each estimated deployment model and each scenario via Monte Carlo simulations.…”

Cooperative Infrastructure and Spectrum Sharing in Heterogeneous Mobile Networks

“…While we address a greenfielddeployment of small cell BSs, [31] assumes operators will pool together their existing macro-cell RAN networks and make joint decisions on future changes to their aggregated RAN such as decommissioning, upgrading or adding new sites. Also in [32], a recent work by Kibilda et al, the shared network is created by pooling together the operators' individual network infrastructure and/or their respective licensed spectrum. This work compares the gains from infrastructure and spectrum sharing when adopted separately and combined (full sharing) on the basis of classical performance indicators such as throughput and coverage probability obtained by means of stochastic geometry models; such gains are shown to strongly depend on the spatial correlation of the individual network deployments and densities while infrastructure and spectrum sharing gains do not sum up as full sharing introduces a tradeoff between data rate and coverage.…”