Yann Gripay, LIRIS Lyon,

Monday April 26th, 10:00a.m. Salle Aurigny (D165)

A Data-oriented Declarative Approach for Pervasive Environments

Pervasive environments pose new challenges in order to exploit their full potential, in particular through the management of complex interactions between distributed resources. Their heterogeneous data sources and functionalities are not homogeneously manageable in today’s systems. This is a big issue when building pervasive applications: imperative programming languages (e.g., C++, Java), classical query languages for databases (e.g., SQL), and network protocols (e.g., JMX, UPnP) must be combined in ad hoc developments, which is neither convenient nor suitable as a long-term solution.

Declarative approaches offer the advantage of providing a logical view on resources that abstracts physical access issues and enables optimization techniques. SQL queries over relational databases are a typical and well-known illustration of those approaches. Therefore, querying data sources and functionalities in a declarative way is recognized as a major issue in pervasive environments in order to simplify the development of applications. Despite a lot of propositions, a clear understanding of the interplays between relational data, data streams and services is still lacking and is the major bottleneck toward the declarative definition of pervasive applications, instead of the current ad hoc development of such applications.

In our work, we propose a data model for pervasive environments, namely the SoCQ data model (standing for Service-oriented Continuous Query).We define a data-oriented view of those environments: the standard notion of database is extended to come up with a broader notion, defined as relational pervasive environment, integrating in the same model relational data, data  streams and distributed services. We also define an algebra, namely the Serena algebra, that enables the declarative expression of one-shot or continuous queries homogeneously combining those data sources. We moreover propose a SQL-like language based on this algebra. A prototype of Pervasive Environment Management System (PEMS) that supports our data model has been implemented using the OSGi framework, and experimentations have been conducted to validate our approach.

The perspectives of this work focus on query optimization issues in this new context and on the integration of this approach within P2P environments. Applications are envisioned in the domain of intelligent buildings (exploiting P2P sensor networks) and in the domain of social networks (information sharing at large scale).