Adaptive Top-k Query Processing in Intermittently Connected Networks

This paper addresses the efficient processing of non-blocking top-k queries in intermittently connected networks, which is reporting the k largest results according to a user-specified ranking function by pipeline model. Top-k query algorithms for intermittently connected networks suffer from how to get steady performance via unpredictable, slow, or bursty network traffic. In this paper, we propose a new adaptive top-k query algorithm, called adaptive and progressive ranking (APR). APR accesses remote sources asynchronously under the server push model, and needs not any user-predefined parameter. Theoretical analyses results show that, compared with the previous approaches, APR is a more efficient solution for producing fast query responses in intermittently connected networks.