Towards a Decentralized Search Architecture for the Web and P2P Systems

Search engines are among the most important applications or services on the web. Most existing successful search engines use a centralized architecture and global ranking algorithms to generate the ranking of documents crawled in their databases, for example, Google's PageRank. However, global ranking of documents has two potential problems: high computation cost, and potentially poor rankings. Both of the problems are related to the centralized computation paradigm. We propose a decentralized architecture to solve the problem in a P2P fashion. We identify three sub-problems in the big picture: a logical framework for ranking computation, an efficient way of computing dynamic local ranking, and a cooperative approach that bridges distributed local rankings and collective global ranking. In the paper we summarize the current knowledge and existing solutions for distributed IR systems, and present our new ideas. We also provide initial results, demonstrating that the use of such an architecture can ameliorate the above-mentioned problems for Web and P2P search engines.1