Transactional operations in sensor environments – TransCap

Sensor querying has become a very popular research topic in many areas of computer science such as networks, databases, embedded operating systems and more generally, ubiquitous computing. This is due to the fact that the contributions of these tiny yet intelligent devices to the new "information everywhere" paradigm are now well recognized. Traditionally, sensor querying and management had been made in an application dependant way. Each sensor application had its own sensors querying, deployment and configuration scripts, and its own sensor data processing techniques. With the multitude of "sensor-aided" applications, a need for a more generic, reusable way of sensor querying had risen. Sensor database systems (SDBS) attempt to fulfill this need, and to provide declarative query formulating and processing techniques. In fact, the reason of this evolution is similar to that of emergence of DBMS a couple of decades ago, i.e. transition from application dependant data files to the application independent databases. However, in spite of the fact that reliable data processing is a core functionality of data management systems, very few coherency issues have been explored for SDBS and few works concerning the quality of service in sensor data management is reported. Most projects on sensor querying suppose that sensor farms characteristics do not change during query evaluation and consider that sensor behavior is stable. In other words, they suppose that modification of the sensor farms state do not impact continuous query evaluation. Nevertheless, with increasing capacity and decreasing cost of sensors, large-scale heterogeneous sensor systems are appearing. We strongly believe that next generation of these systems will include, besides continuous queries on sensor data, system management operations that may require some transactional properties. It is the fact that, in a large and heterogeneous sensor system, it should be possible to concurrently support management operations and continuous read-only queries. For example, adding/removing sensors in the farm or updating sensors firmware may leads to inconsistency for concurrent read-only queries. In other words, it is necessary to consider the impact of management operations on continuous read-only queries. Since management operation is a modification of the sensor farm state, requirements on concurrency support, atomicity and more generally on some forms of transactional properties appear. The aim of this project is to achieve significant breakthroughs in the study of the following challenging points: * To identify the exact characteristics of concurrent operations needed in sensor contexts. * To characterize the transactional properties required for management of sensor systems. * To propose appropriate protocols and services to fulfill mentioned transactional requirements. The overall goal is to define a sensor database 'management' system (SDBMS), which guarantees appropriate degree of transactional properties.