Attribute-based data access over p2p systems
Abhishek Gupta, Divyakant Agrawal, Amr El Abbadi · 2004
Peer-to-peer (P2P) systems provide a distributed and scalable alternative to access services and data over the network which is in contrast to the traditional client-server model. The basic idea in these systems is to harness the collective storage resources available at the large number of peers, obviating the need for a few servers with large storage. In recent years, P2P computing has emerged as a powerful distributed computing model for large scale systems. P2P systems have been widely used for large scale file sharing and instant messaging applications. The initial P2P designs could either not scale to match their popularity or caused a lot of overhead on the networks. More recently Distributed Hash Table based structured designs for P2P systems have been proposed that exhibit excellent scalability and routing and lookup performance. These designs are however limited in the kind of applications they can support because of the simple hash table interface they provide. DHT systems simply rely on the names of the data objects to distribute, index and locate them. In this thesis, we present designs of DHT based P2P systems that extend the interface to support attribute based data access by extracting semantic information from the queries. We describe solutions for approximate as well as exact answers to range queries. We apply the designs to support Select, Project, and Join based SQL queries and also present an application to Publish/Subscribe systems. We have also developed a prototype system for the proposed Publish/Subscribe system and evaluated its performance on PlanetLab machines distributed across the world.