NOVA is a state-of-the-art NVM-based file system that ensures the consistency of data and metadata in the event of power outage through its log-structured design. NOVA implements per-inode logging for metadata and uses per-core data structures such as inode table, journaling space, and memory allocator to boost performance by supporting concurrent I/Os. However, NOVA does not fully exploit the scalability of I/O throughput when multiple I/Os are being performed to shared files in a manycore server environment. Also, NOVA does not apply space optimization techniques such as deduplication or compression to make NVM space efficient.

In this talk, I will first introduce various optimization techniques, such as fine-grained locking and lock-free data structures for I/O scalability of NOVA. Second, I will present our recent work on space optimization in NOVA, an offline deduplication method with DRAM-free persistent deduplication metadata.

Youngjae Kim is an Associate Professor of Computer Science and Engineering Department at Sogang University. Before joining Sogang in 2016, he was a research staff scientist at US Department of Energy’s Oak Ridge National Laboratory. He received his Ph.D. in Computer Science and Engineering from Pennsylvania State University in 2009, his M.S. in Computer Science from KAIST, and his B.S. in Computer Science from Sogang University. His research interests include file and storage system, database system, concurrency, and parallel processing, and distributed system.

If you want to get more information about the colloquium, please refer to colloquium page: https://cs.kaist.ac.kr/colloquium/.