Cost-Based Storage of the R-tree Aggregated Values over Flash Memory

The flash memory due to its shock resistance, power economy and non-volatile nature is considered as a very popular storage device. It is widely used in mobile phones, sensor networks and hand-held devices. What attracts the attention is the data storage in the flash memory. Database vendors try to adapt their systems to the flash memory architecture. To make it efficient, database indexes, such as: B-tree or R-tree, must be implemented from the scratch in order to take into account the flash memory limitations. The R-tree is wildly used in the geographical databases for indexing the spatial objects. The aggregated R-tree (aR-tree) extends the original R-tree with the aggregated values, which are connected with the nodes. In this paper, we deal with the materialization technique of the aggregated values in the aR-tree index. We claim that not all the aggregated values should be explicitly stored in the flash memory. We propose a Cost-Based Method which chooses the most beneficial set of the aggregated values for materialization. Our method is particularly beneficial for devices with limited capacity and is optimized for the flash memory architecture.