KEYNOTE: Keyword search by node selection for text retrieval on DHT-based P2P networks

Efficient full-text keyword search remains a challenging problem in P2P systems. Most of the traditional keyword search systems on DHT overlay networks perform the join operation of keywords at document level, which consumes a huge amount of storage and bandwidth. In this paper, we present KEYNOTE, a novel keyword search system that performs the join operation at node level. Compared to the traditional keyword search systems on DHTs, KEYNOTE can greatly reduce the storage and communication cost. To forward a query to the relevant nodes for searching documents, two effective node selection methods are presented. To address the hot spot problem in Chord overlay networks, an efficient load balancing scheme is introduced. Simulated experimental evaluation with up to 8,000 nodes and over 600,000 real-world documents validates the practicality of the proposed system.