Data management in human networks
Calicrates Policroniades, Pablo Vidales, Martin Roth, Daniel Kreienbühl · 2007
In this paper we study the use of a semantically rich storage model to fulfill the data transmission requirements of challenged networking environments, which are characterised by long delays and frequent communication disruptions. Practical experience shows us that the highly successful data abstractions of mainstream storage systems (e.g. monolithic file representation) operate poorly in emergent networking environments such as Delay Tolerant Networks (DTNs); short contact times do not allow for complete file or bundle transmissions. We have ported and integrated two systems in order to provide a solution that overcomes many of the data transmission challenges of DTNs: a semantically rich storage system (Datom) and a network framework capable of exploiting this augmented expressive power (Haggle). Our solution, Bedouin, enables both systems to run on resource-constrained devices. It facilitates meaningful data exchanges in challenged networks supporting the principle of infrastructure-independent networking, and exploiting human mobility and opportunistic connectivity. The design and function of a proof-of-concept Bedouin-based peer-to-peer file sharing application for human networks, called Caravan, is included. Experimental results demonstrate that our solution enables applications to work correctly in spite of intermittent data exchanges and disruptions while maximising the amount of useful data delivered to applications.