NAND flash-based devices are widely used in many computer systems, ranging from embedded systems to big-data analytics systems. Architecting storage systems suitable for NAND devices is critical for improving I/O performance and device longevity. In this talk, I present a BlueDBM system, a flash-based storage appliance for big data analytics, which is being built by CSAIL MIT. In particular, I focus on the design and implementation of a software platform for BlueDBM, which includes three novel features: in-storage processing, refactored I/O stacks, and storage-to-storage networks. I first explain an in-storage processing (ISP) framework of BlueDBM that accelerates data analysis by enabling easy use of hardware accelerators implemented in FPGA inside the storage device. I also present the reduced I/O system architecture of BlueDBM that accomplishes short I/O response times and high I/O throughput by refactoring storage I/O stacks, including a file system, a device driver, and stora ge firmware. Finally, I show preliminary results for clustered BlueDBM nodes that exploit high-speed storage-to-storage networks.

Sungjin Lee received the BE degree in electrical engineering from Korea University in 2005 and the MS and PhD degrees in computer science and engineering from the Seoul National University in 2007 and 2013, respectively. He is currently working as a postdoctoral associate in the Computer Science and Artificial Intelligence Laboratory at the Massachusetts Institute of Technology. His current research interests include storage systems, operating systems, and embedded software.