Practical data-centric storage

Abstract

Most data retrieval mechanisms in wireless sensor networks adopt a data-centric approach, in which data is identified directly by name rather than through the location of the node on which it is stored. Initial data-centric methods, such as directed diffusion and TinyDB/TAG, focused on the conveyance of data. One of the advantages of these algorithms is that they do not require point-to-point routing, which has proved to be difficult and costly to implement in wireless sensor networks, and instead require only the simpler and more robust tree-construction primitives. Some recent data retrieval proposals have extended the data-centric paradigm to storage. Data-centric storage uses in-network placement of data to increase the efficiency of data retrieval in certain circumstances. Unfortunately, all such proposals have been based on point-to-point routing, and therefore have faced a significant deployment barrier. In this paper we hope to make data-centric storage more practical by removing the need for point-to-point routing. To that end, we propose pathDCS, an approach to data-centric storage that requires only standard tree construction algorithms, a primitive already available in many real-world deployments. We describe the design and implementation of pathDCS and evaluate its performance through both high-level and packet-level simulations, as well as through experiments on a sensor testbed.