CTR

Abstract

Top-k queries, which retrieve the k data items with the highest scores, have been receiving considerable attention because of their utility for many real life applications. In this paper, we propose an efficient query routing scheme enabling top-k data retrieval by only necessary nodes, named CTR. CTR employs a new clustering framework for top-k query processing, and in this framework, nodes holding high rank data items become ClusterHeads (CHs), and top-k queries are transmitted between CHs via gateway nodes which belong to multiple clusters. Each CH maintains a record of hop counts between itself and high rank data items, enabling it to judge whether or not to transmit a query on the fly. As a result, CTR reduces overhead while keeping high accuracy of the query result. Extensive experiments have demonstrated that CTR functions well in terms of accuracy of the query result, overhead, and delay.