A Small File Merging Strategy for Spatiotemporal Data in Smart Health

Abstract

With the rapid development of smart health, the health sensors and wearable devices bring huge amounts of small files of spatiotemporal data, which are distributed in different servers, and affect the I/O performance of the system seriously. There are many methods to solve the problems of a small file, but most of them are applied to specific applications. However, due to the influence of user access behavior and data type, these methods are not effective when applied to sensors data in smart health. In this paper, a novel small file merging strategy for smart health is proposed. By analyzing the features of health sensors data and the preferences of user access, the strategy uses spatiotemporal clustering for the historical user access information. Then, weights are applied to these clusters based on the access density to determine the access-related spatiotemporal range. Finally, the spatiotemporal range is used to merge small files. The experimental results show that the merging strategy is simple but efficient, and it can effectively reduce user access delay for small files of spatiotemporal data in smart health.