Caching in Ultra-Dense Small Cell Networks with Limited Backhaul

In this paper, we investigate the influence of limited backhaul on the performance of caching enabled ultra-dense small cell networks (USCN). In particular, an analytical framework has been presented to evaluate the performance of USCN in terms of area spectral efficiency (ASE). It is shown that the constraint of backhaul capacity would significantly influence the performance of content caching and retrieving in USCN. For instance, it is shown that, if the resulting interference is not properly handled, caching more content would result in a decrease in ASE of USCN in the unlimited backhaul regime. This contradicts with the limited backhaul regime, in which network ASE could be notably improved if more content is pre-fetched. Moreover, it is shown that network ASE would experience a faster diminish as well if more content is retrieved via backhaul. The reason is that USCN turns from interference-limited into backhaul- limited regime with the growing BS and user densities. On this account, we have designed a probabilistic content retrieving strategy to relieve the bottleneck brought by limited backhaul. Targeting at maximizing the ASE of USCN, we further optimize the content retrieving probability (CRP). With the derived sub-optimal CRP, it is shown that the scaling law of network ASE (with varying BS density) could be fundamentally improved. Therefore, the results of this work could provide insight on the application of caching in USCN with limited backhaul constraint.