Efficient non-blocking top-k query processing in distributed networks

Incremental access can be essential for top-k queries, as users often want to sift through top answers until satisfied. In this paper, we propose the progressive rank (PR, for short) algorithm, a new non-blocking top-k query algorithm that deals with data items from remote sources via unpredictable, slow, or bursty network traffic. By accessing remote sources asynchronously and scheduling background processing reactively, PR hides intermittent delays in data arrival and produces the first few results quickly. Experiments results show that PR is an effective solution for producing fast query responses in the presence of slow and bursty remote sources, and can be scaled well.