Geodupa - a geography-driven partial membership algorithm

Abstract

The popularity boost of mobile devices, as well as their technological advances, are fostering the development of new application paradigms. One of which, designated as Participatory Sensing, is embedded with a strong community philosophy, in the sense that it relies in users sharing and contributing with data. By gathering, processing and sharing contextual data, new and interesting applications are possible, such as the monitoring of vehicle traffic or road conservation. To enable these applications, it is important to have an underlying communication infrastructure that allows users to exchange information efficiently. Users of Participatory Sensing applications deal, most frequently, with information related to their close physical surroundings. Based on this premise, this dissertation presents a decentralized membership substrate that restrains node visibility to geographical neighborhoods as a way to improve communication performance. To that end, the proposed algorithm divides the user network into two hierarchical levels. The higher-level is managed by an existing one-hop DHT and its participants are organized to exploit the partitioning of the physical space. The lower-level is composed by the groups of nodes associated to each region. An experimental evaluation has revealed that it is capable of achieving lower communication costs when compared to a full-membership solution.