Efficient querying of distributed sources in a mobile environment through source indexing and caching
Mobile devices have become a part of the everyday life; they are used anywhere and at anytime, for communication, looking up information, consulting an agenda, making notes, playing games etc. At the same time, the hardware of these devices has evolved significantly: e.g., faster processors, larger memory and improved connectivity … The hardware evolution along with the recent advancements of identification techniques, has lead to new opportunities for developers of mobile applications: mobile applications can be aware of their environment and the objects in it. Combining these new opportunities with the Web, allows mobile applications to use services and information of nearby objects (e.g., a mobile application that informs you of the restaurants that are nearby your current position without the need to enter your current position). The SCOUT framework, currently being developed at the WISE lab, supports the development of context- and environment-aware mobile applications, by providing a conceptual and integrated view on the environment called the Environment Model. This Model comprises metadata on physical entities found nearby the user, and the user’s own profile information, and thus allows applications to become aware of (and responsive to) the user’s physical environment and context. SCOUT is a decentralized and distributed solution, where no single centralized server is required for storing context-sensitive data or integrating data from various sources. Instead, this integration is achieved via the locally maintained Environment Model, while each content provider is responsible for making available and managing their own data. In order to facilitate information integration across different heterogeneous sources, Semantic Web technology is employed. Until now, the Environment Model has been stored as a fully materialized view; in other words, all of the data (i.e., encountered data sources) is kept locally. This thesis investigates how the Environment Model can be constructed and managed more efficiently. Each of the strategies mentioned below have been tested extensively in different scenarios, in order to determine the most suitable ones and to indicate where there is room for improvement. Firstly, we have investigated several ways of storing summary information on encountered data sources to determine which sources contain relevant information for a given query. The goal of these strategies is to avoid having to include all encountered sources when answering a query. Secondly, we have developed several caching strategies where some data from encountered sources is kept locally, to avoid having to download a relevant sources (as identified by one of the strategies mentioned above) every time it is needed to solve a query issued to the Environment Model.