—The recently proposed Pocket Switched Network paradigm takes advantage of human social contacts to opportunistically create data paths over time. We examine how effective such a network can be, and how to create effective paths in such networks. Using a break-and-observe approach, we establish that time correlations and the contact occurrence distribution primarily determine long-term delivery ratio evolution. In contrast, shorter time windows exhibit significant variation in performance, and the formation of a large clique core during the window is correlated with the fraction of data delivered, as well as the speed of delivery. We next examine paths that successfully deliver data and find that although their delivery times vary widely, a surprisingly simple randomised scheme can pick nearoptimal paths. This scheme is also robust in the sense that it distributes load evenly, and renders the network resilient against failures of critical nodes. Taken together, our findings sugges...
Nishanth Sastry, Karen R. Sollins, Jon Crowcroft