Dynamic selection of indexes and views materialize with algorithm Knapsack

The volume of information contained in the databases is constantly increasing, the spatio-temporal data collected by the sensor network form a set of thematic, integrated, nonvolatile and historical data organized for use in helping to decision. Typically, this process is performed using temporal, spatial, and time-space queries. This in turn increases the execution time of decision requests. To optimize the load of queries, it is essential to choose the multiple selection of similar structures. In the literature, several optimization techniques have been identified: materialized views, indexes, fragmentation, planning and buffer management. These techniques do not consider the change in the request load and the data change. In this article, we propose an optimal so-called dynamic selection based on the use of indexes and materialized views. a solution is said to be optimal when it answers all the requests with a minimum response time. Our approach allows to optimize the execution of the requests and to adapt an optimal solution to the modifications of data in the base one applies the algorithm of backpack.