| In recent years, many Peer-to-Peer (P2P) systems have emerged and attracted millions of users to share files, storage and other resources on the Internet. To provide efficient routing on such large-scale networks, many structured P2P overlays were proposed. However, these systems provide only a basic name lookup service. This dissertation presents a suite of mechanisms to support complex queries on top of structured P2P networks. First, to support keyword-based search, we developed Keyword Fusion that not only reduces query latency and traffic, but also balances the storage consumption and query processing load among peers. Second, to support range queries, we developed a novel Range Hash Tree (RHT) that processes a range query with high predictability and scalability having its query processing time independent of both the queried range size and the number of total matching data objects. Finally, to help users select a more reliable query result from various peers, we developed a new reputation management system called R-Chain where each peer maintains its own transaction history locally as reputation. R-Chain guarantees the availability of reputation information for active peers with only limited network overhead for reputation maintenance. Experiments show that the proposed mechanisms are lightweight, and highly efficient in supporting complex queries. |
