Enhanced database support for location-based services

Abstract

The ubiquity of GPS-enabled mobile devices and sensors have led to the explosive growth of time-stamped location data. Consequently Location-Based Services (LBS) has become a popular technology impacting various aspects of our lives. LBS applications are characterized by very high rate of location record updates, and many concurrent historic, present and predictive queries. Commercial LBS providers rely on relational databases to manage their data. However, traditional relational databases do not provide adequate support to meet the growing demands of many LBS systems. Moreover, existing indexing techniques that support historical queries are unable to sustain high update and query throughput as required by many LBS applications. To address this, we propose to exploit in-memory database techniques and present a few key ideas to support high performance commercial LBS. We also introduce a novel in-memory spatio-temporal index in which the spatial domain is organized as grid cells and for each grid cell partial temporal indexes are maintained for moving objects that visited the cell. The partial temporal indexes are implemented as compressed bitmaps. Using fast bitmap operations and utilizing parallelism rendered by multi-core systems, our system offers significantly better performance than traditional relational databases.