Paper Title: Increasing Database Performance by Placing Transactional Functions in an SSD

Paper Abstract: Current DBMS and file systems use transaction logging and journaling for metadata to support transactional atomicity. Unfortunately, this approach causes duplicated writes and multiple versions of data, leading to more overall write amplification and lower performance. One way to reduce the degradation is to handle transaction logging and journaling in an SSD. The result should be improved DBMS performance and reduced write amplification (leading to longer SSD lifetimes). Applying this approach to an in-memory NoSQL DBMS illustrates the improvements that can be obtained.

Paper Author: Woo Suk Chung, Director, SK hynix

Author Bio: Woo Suk Chung is the team lead for Advanced Software Solutions at SK Hynix Memory System R&D. His research focuses on future memory/storage solutions, and he investigates how next generation technology will serve the needs of enterprise SSDs. Before joining SK Hynix, he was a project leader in the Mobile and Wireless Group at Broadcom. While leading the Android system development team there, he received the key achievement award and the best performance award for the successful implementation of a low power, high performance application processor. Before that, he was a Sr Engineer at Samsung System LSI where he developed mobile solution for WinCE and Linux for the Samsung Application Processor. Mr. Chung holds 18 patents and a BSEE and MSEE from Hanyang University in Seoul, Korea.