Implementing Big Data Algorithms on GPUs

Abstract

Algorithms for processing large, unstructured data sets have shown great promise in implementations on modern graphics processors (GPUs), with many implementations reporting 20-70x speedup over comparable CPU-only versions of the same algorithms. In this senior project research, our goal is to implement an efficient, highly scalable SQLite database on GPU, test an optimized implementation of a data sorting algorithm like GPU-Quicksort, and demonstrate the speed potential of GPU-enhanced computation on a typical big-data search and aggregation algorithm like MapReduce.