View Position
| Project Title | HPC database implementation intern |
| Summary | The primary objective of this proposal is to develop a curriculum module focused on improving the speed of database queries based on the MapReduce paradigm using multiple GPUs. Previous work in this area has shown significant speedups in SQL queries using a single GPU system. One implementation uses a custom data structure to partition the table into fixed-sized groups of rows that allow the queries to run in parallel. The intern will first learn how SELECTs and JOINs work and then extend the previous data structure for use on the Blue Waters multi-GPU system. Finally, the intern will test these parallel queries with very large data sets and compare their performance with traditional implementations. |
| Job Description | The intern will work with a local mentor to (1) learn about relational database queries, specifically SELECTs and JOINs and their implementations, (2) write programs in CUDA C to parallelize the queries and ensure their accuracies (3) update the code to run on the Blue Waters system and test performance improvement versus scalability (4) document a manual for the curriculum module. |
| Conditions/Qualifications | Must have programming experiences in C/C++, preferably in CUDA and an interest in learning about relational databases. |
| Start Date | 05/31/2015 |
| End Date | 05/31/2016 |
| Location | Department of Mathematics and Computer Science Hampden-Sydney College Hampden-Sydney, VA 23943 |
| Interns | Linh Nguyen
|