On Designing Archiving Policies for Evolving RDF Datasets on the Web

Abstract

When dealing with dynamically evolving datasets, users are often interested in the state of affairs on previous versions of the dataset, and would like to execute queries on such previous versions, as well as queries that compare the state of affairs across different versions. This is especially true for datasets stored in the Web, where the interlinking aspect, combined with the lack of central control, do not allow synchronized evolution of interlinked datasets. To address this requirement the obvious solution is to store all previous versions, but this could quickly increase the space requirements; an alternative solution is to store adequate deltas between versions, which are generally smaller, but this would create the overhead of generating versions at query time. This paper studies the trade-offs involved in these approaches, in the context of archiving dynamic RDF datasets over the Web. Our main message is that a hybrid policy would work better than any of the above approaches, and describe our proposed methodology for establishing a cost model that would allow determining when each of the two standard methods (version-based or delta-based storage) should be used in the context of a hybrid policy.