Decentralized Indexing over a Network of RDF Peers

Despite the prospect of a vast Web of interlinked data, the Semantic Web today mostly fails to meet its potential. One of the main problems it faces is rooted in its current architecture, which totally relies on the availability of the servers providing access to the data. These servers are subject to failures, which often results in situations where some data is unavailable. Recent advances have proposed decentralized peer-to-peer based architectures to alleviate this problem. However, for query processing these approaches mostly rely on flooding, a standard technique for peer-to-peer systems, which can easily result in very high network traffic and hence cause high query response times. To still enable efficient query processing in such networks, this paper proposes two indexing schemes, which in a decentralized fashion aim at efficiently finding nodes with relevant data for a given query: Locational Indexes and Prefix-Partitioned Bloom Filters. Our experiments show that such indexing schemes are able to considerably speed up query processing times compared to existing approaches.