Approximate query answering in unstructured peer-to-peer databases

Peer-to-peer networks are considered the new generation of distributed databases, termed peer-to-peer databases. With very large size, open architecture, and extreme dynamism and autonomy, approximate query answering is arguably the most promising approach for query answering in peer-to-peer databases. To enable approximate query answering, in this dissertation we propose a set of universal sampling operations specifically designed for probing the data in peer-to-peer databases. We complement these operations at the bottom tier by a set of approximate query processing techniques at the top tier to develop a two-tier system for answering both set-valued and aggregate queries in peer-to-peer databases.