Experimental Study of Exploiting GPU for Database Query Optimization

Alfattani Sami(0951207)

With the extremeness of the computer revolutions, the amount of data had being swelling day by another. Today the researchers are looking for methods, techniques and solutions to reduce the execution time as much as possible. That means the increasing of performance with saving time, and energy. Parallel computing is considering one of the good ideas to increase the query execution performance. One of the interesting devices that supporting parallel computation is Graphic processing Unit (GPU). In this study, we will try to find a model that makes a database query can be optimized using GPU. Of course there is no guarantee of that the GPU will process data always quickly, but in some cases and under some circumstances GPU can be reliable way to optimize a query. The objective of this research is to answer two questions. First, when (in which data size) the GPU can be faster than CPU in executing a database operation? Second, given a query requires more than one operation to extract the final result, which operation can be executed on GPU and the query will be optimized overall?