Biography

Dr. Myoungsoo Jung is Assistant Professor at Yonsei University. Dr. Jung earned his Ph.D. in Computer Science at Pennsylvania State University and his M.S. in Computer Science from Georgia Institute of Technology, and an M.S. in Embedded System from Korea University in Seoul. Dr. Jung has many years of industry experience, several industrial U.S. patents related to multi-channel SSDs, and approximately sixty technical papers regarding SSD flash firmware and kernel-level file systems. His research has been nominated as best paper from the Institute of Electrical and Electronics Engineers/Association for Computing Machinery (IEEE/ACM) Internal Conference for High Performance Computing, Networking, Storage and Analysis 2013 (SC’13). He received core grant awards from National Science Foundation (NSF) and Department of Energy (DOE), respectively, and the Lawrence Berkeley National Laboratory Award (LBNL) of Excellence. His current research interests include coprocessor architecture (e.g., MIC/GPU), FPGA-based accelerators, advanced computer architecture, and operating systems on emerging non-volatile memory and solid state drive technologies

Abstract

Thanks to massive parallelism in modern coprocessors such as Graphics Processing Units (GPUs) and Many Integrated Core (MIC), emerging data processing applications in accelerator computing exhibit ten-fold speedups compared to CPU-only systems. However, this coprocessor-based acceleration is limited in many cases by the significant data movement overheads and inefficient memory management for host-side storage accesses. In the meantime, Solid State Disks (SSDs) are undergoing dramatic technological shifts and architectural changes by employing hundreds of NAND flash chips, multiple I/O channels, multiple cores, and high speed interfaces such as PCI Express and SATA 6Gpbs. We will tee this talk up by advocating a radically different approach to the design of next generation accelerators by bringing accelerator resources and storage resources together, and tailoring these to improve data flow ultimately maximizing performance and power efficiency. In addition, this talk will cover six different NVM research topics that span from datacenters to enterprise edge to connected devices.