Reducing expenses of top-k monitoring in sensor cloud services

Abstract

In sensor cloud services, the expense is charged based on the amount of resource usage, e.g. data requests. This paper originally presents an expense-minimizing framework for top-k monitoring in sensor cloud services where the expense is denoted by the costs of data requests. Instead of fetching all the latest data in each timestamp, we propose a novel ε-top-k query delivering approximate top-k answers with a probabilistic guarantee on the selectively-fetched dataset which is a combination of certain and uncertain data (modelled by their age). In addition, using a cloud environment as well as our proposed method to process ε-top-k queries can alleviate the computing-intensive computations, so it is not only cheaper but even faster than an ordinary top-k calculation method. The extensive experiments on the real-world climate datasets demonstrate that our methods can reduce the expense by more than half with desirable accuracy.