Accelerating relational database operations using both CPU and GPU co-processor

Data is evolving and the number of existing data sources is vastly growing. Therefore, there is a compelling need for effective techniques to store, retrieve and process such massive data. Significant speed-ups at a small cost can be achieved by deploying co-processors such as GPUs. To this end, in this paper, we propose a new hybrid query processing technique that makes use of the capabilities of CPUs and GPUs. The proposed approach breaks down each SQL statement into smaller parts during the parsing process. It then automatically manages the distribution of different query parts to be executed either on the CPU or parallel on the GPU and CPU. To achieve this, we developed and implemented the proposed approach on a SQL server database using the .Net framework instead of working under the Linux environment. The performance of the proposed approach is validated using different workloads and the results demonstrate that the proposed GPU-based query processor achieved speedup up to 39 as fast as multi-core CPUs.