Performance Predictability for Flash-Based Storage

Achieving predictable performance for flash-based storage is becoming increasingly

important in this Age of Big Data. However, this is no easy task as the underlying flash

memory characteristics continue to deteriorate, requiring an understanding of hardware

trends, optimizations across layers, and designs with scalability and adaptability in mind. In

this talk, I will present an autonomic approach for scheduling SSD-internal tasks by

dynamically managing their progress, and examine the design tradeoffs of existing SSD’s

reliability enhancement techniques. Lastly, I will discuss and outline on-going future research

directions beyond these studies.

Biography:

Bryan S. Kim is a postdoctoral researcher in the Department of Computer Science &

Engineering (CSE) at Seoul National University (SNU). He received his B.S. degree in

Electrical Engineering and Computer Science (EECS) from UC Berkeley in 2006, and his

M.S. in EECS and Ph.D. degree in CSE from SNU in 2009 and 2018, respectively. Prior to

his Ph.D. studies, he was with the Storage Technology Lab at the R&D division of SK

Telecom, the largest telecommunication operator in South Korea. His research interests

span across the storage stack from the underlying medium to the system software, with

particular emphasis on flash-based systems and devices. His recent publications include

papers at RTAS, USENIX ATC, FAST, and SIGMETRICS.