Making the most of your contacts: transfer ordering in data-centric opportunistic networks

Opportunistic networks use unpredictable and time-limited contacts to disseminate data. Therefore, it is important that protocols transfer useful data when contacts do occur. Specifically, in a data-centric network, nodes benefit from receiving data relevant to their interests. To this end, we study five strategies to select and order the data to be exchanged during a limited contact, and measure their ability to promptly and efficiently deliver highly relevant data. Our trace-driven experiments on an emulation testbed suggest that nodes benefit in the short-term from ordering data transfers to satisfy local interests. However, this can lead to suboptimal long-term system performance. Restricting sharing based on matching nodes' interests can lead to segregation of the network, and limit useful dissemination of data. A non-local understanding of other nodes' interests is necessary to effectively move data across the network. If ordering of transfers for data relevance is not explicitly considered performance is comparable to random, which limits the delivery of individually relevant data.