Abstract
Consider a system in which there are N nodes. Each node is characterized by a set of attributes. The attribute values may change along time. The problem is to answer queries which look for nodes with attributes satisfying criteria such as range of values or similarity. In peer-to-peer (P2P) systems, this problem is often solved by storing the attribute values in a distributed hash table supported by a structured P2P system. Unfortunately, the hash-based solution makes it difficult to answer queries such as range and k-nearest neighbor queries. In this study, we propose to concentrate on the attributes of the peers and use peer interconnection to exploit their relationship. The result is an attribute-based overlay network (AON) among the peers to index the continuously changing peer attributes. With AON, two peers with similar attribute values will be close to each other in the overlay. By preserving the attribute relationship of peers in the interconnection, AON can support range queries and k-nearest neighbor queries efficiently. In this thesis, we focus primarily on the feasibility of using AON to support attribute-based queries in P2P systems, by contrasting it with DHT-based overlay in terms of performance and cost under different levels of churns and attribute-changing rates. The simulation results show that AON outperforms DHT-based systems in most cases.