Subscription-based data aggregation techniques for top-k monitoring queries

With the increase of data generation in distributed fashions such as peer-to-peer systems and sensor networks, top-k query processing which is a way to aggregate only a small set of data that sufficiently satisfies many users' preferences, becomes a substantial issue. When data are periodically updated in each epoch e.g., weather information, without any techniques, a naive solution is to aggregate all data and their updates to ensure the completeness of final answers, however, it is too costly in terms of data transfer especially for data aggregator nodes (intermediate nodes). In this paper, we propose a top-k monitoring query processing method in 2-tier distributed systems based on a publish-subscribe scheme. A set of top-k subscriptions specifying summary scope of users' interests is informed to aggregators to limit the number of transferred data records for each epoch. In addition, instead of issuing subscriptions of all queries, our method identifies a small set of minimal subscriptions as well as utilizes some adaptive heuristic rules to efficiently maintain those subscriptions resulting in lower communication overhead. Our experiments through both synthetic and real datasets show that our technique is efficient and outperforms other comparative reactive methods.