A major challenge in database systems research is obtaining interesting workloads that one can use to evaluate new ideas. Previous benchmark frameworks only support a small number of workloads or a single DBMS. Our OLTP-Bench framework improves on these earlier efforts by supporting a greater selection of both benchmarks and DBMSs, but it is still limited.
Given this, we are developing a catalog of database workloads from open source code repositories. We will create a tool that automatically searches the Internet for database applications and runs them locally in order to learn how they use a DBMS. We will develop techniques for approximating the data and workload properties, and automatically create the corresponding synthetic generators that can be used in our benchmark framework.
This research was funded (in part) by the National Science Foundation (III-1423210).Visit Project Homepage