LinkedPeers: A distributed system for interlinking multidimensional data

Abstract

In this paper we present LinkedPeers, a distributed system designed for efficient distribution and processing of multidimensional hierarchical data over a Peer-to-Peer overlay. he system design aims at incorporating two important features, namely large-scale support for partially-structured data and high-performance, distributed query processing including multiple aggregates. To achieve that, LinkedPeers utilizes a conceptual chain of DHT rings that stores data in a hierarchy-preserving manner and is able to adjust both the granularity of indexing and the amount of pre-computation according to the incoming workload. Extensive experiments prove that our system is very efficient achieving over 85% precision in answering queries while minimizing communication cost and adapting its indexing to the incoming queries. © 2011 Springer-Verlag Berlin Heidelberg.