Efficient Model-driven Query Processing Based on Data Regeneration

Nowadays, sensor data is generated in large amounts. Stor- ing or transmitting all the sensor’s measurements might not be the ideal choice because of the volume (and rate) at which it is generated. But we also cannot easily discard it, since ev- ery data might contain relevant information and be required while processing queries. One approach to avoid storing and transmitting large amounts of data is to create a sufficiently accurate model over the data and then store or transmit that model. Later, at the time of query processing, the straight forward approach is to regenerate the required data from the model(s) and proceed with query processing, or answer the queries directly on the model(s). We ask the question of how computationally ex- pensive are various query processing techniques, given the availability of the models only.