Search

Cellular operators are considering large-scale small cell deployment in urban traffic hot spots to combat the looming capacity crisis in their networks. However, connecting a large number of small cells to the network core using current backhaul technology is costly. In this paper, we propose message ferrying as a low-cost backhaul solution that uses mobile phones of vehicle occupants as an army of ferries to transfer network data between small cell base stations and nearby switching centres. There are a number of challenges in the design and deployment of such a system. Use of phone memory for network data transfer without interfering with user applications is one such challenge that we study in this paper. Although the memory capacity in mobile phones has increased significantly over the years, it is considered a volatile ferrying resource as it is claimed and released dynamically by user applications. If message ferrying is to be transparent to the user, base stations must be able to predict memory availability over a sufficiently long horizon to ensure that data carried in the phone do not get wiped out during transit due to dynamic memory claim by user applications. We experiment with real phones by logging memory availability for extended hours under different usage scenarios. By applying autoregressive models on memory usage traces, we are able to predict the minimum memory available over a 7.5-minute horizon with 94% accuracy. [ABSTRACT FROM AUTHOR]