Skip to Main Content
In today's data-intensive world, the need for data organization has increased dramatically. Distributed systems are dealing with unheard amounts of data arising primarily from the popularization of pervasive computing applications and the so-called “data-in-the-cloud” paradigm. Naturally, agent-coordination systems are affected by this data-increase phenomenon as they are often used as the basis for pervasive-computing frameworks and cloud-computing systems. There have been a few works on coordination system to include data self-organization (e.g. Swarm Linda) however they generally organize their data based on naive approaches where items are either completely similar or dissimilar (1|0 approach for matching of data). Although this approach is useful, in general-purpose systems where the diversity of data items is large, data items will rarely be considered as plainly similar, leading to a situation where data does not self-organize well. In this paper we move towards a general-purpose approach to organization based on an ontology-defined concept relationship in WordNet. In our approach, data items are seen as concepts that have relation to other concepts: tuples are driven towards one-another at rates that are proportional to the strength of tuple relationship. We demonstrate that this approach leads to a good mechanism to self-organize data in data-intensive environments.
Date of Conference: Sept. 27 2010-Oct. 1 2010