A vector model for routing queries in web search engines

This paper proposes a method for reducing the number of search nodes involved in the solution of queries arriving to a Web search engine. The method is applied by the query receptionist machine during situations of sudden peaks in query traffic to reduce the load on the search nodes. The experimental evaluation based on actual traces from users of a major search engine, shows that the proposed method outperforms alternative strategies. This is more evident for systems composed of a large number of search nodes which indicates that the method is also more scalable than the alternative strategies