Summarizing distributed data streams for storage in data warehouses

Data warehouses are increasingly supplied with data produced by a large number of distributed sensors in many applications: medicine, military, road traffic, weather forecast, utilities like electric power suppliers etc. Such data is widely distributed and produced continuously as data streams. The rate at which data is collected at each sensor node affects the communication resources, the bandwidth and/or the computational load at the central server. In this paper, we propose a generic tool for summarizing distributed data streams where the amount of data being collected from each sensor adapts to data characteristics. Experiments done on electric power consumption real data are reported and show the efficiency of the proposed approach.