Structuration de données spatio-temporelles temps-réelles : vers la gestion de la saturation de base de données

The spatio-temporal databases in the monitoring systems of natural phenomena must be able to meet real-time requirements due to the semi-generalized use of sensor networks. The indexing method of the in-memory base must take into account several characteristics: primacy of the most recent data and updates queries, spatial positioning of the sensors, heavy charge load... Our first proposals, the PoTree and the PasTree are designed to index spatio-temporal data issued by respectively fixed and agile sensors. A temporal sub-tree is dedicated to each sensor. A spatial sub-structure, as well as an identifier one for the PasTree are used to differentiate the sensor sub-trees. Database saturation management has led to the development of the StH, based on the PasTree. It uses the notion of heat to sort the data into classes, according to their importance and puts them into escalator like sensor sub-structures. The StH is currently under experimentation. MOTS CLÉS : base de données spatio-temporelles, système d’indexation, contraintes tempsréel, réseau de capteurs, saturation, PoTree, PasTree, StH