SRC25. Parallel Provenance Databases for High Performance Workflows
Student: Jennifer A. Steffens (Drake University)
Supervisor: Justin M. Wozniak (Argonne National Laboratory)
Abstract: In scientific computing, understanding the origins and derivation of data is crucial. Provenance models aim to provide a means of capturing this in an efficient and effective manner. For the Swift/T language, the current provenance handling system requires improvement. In this poster, we discuss the development of a new Swift/T provenance model, the Multiple Parallel Databases Model (MPDM), which parallelizes the real-time storage of provenance data in a user-accessible database system. Utilizing multiple databases in high performance, parallel workflows can increase the practicality of lightweight, relational databases engines such as SQLite, as we show MPDM to be more efficient and have better scalability than the previous, single database model.
Two-page extended abstract: pdf