Center for Research on Intelligent Storage (CRIS) Proposal #1127829
This proposal seeks funding for the Center for Research on Intelligent Storage (CRIS) at the University of Minnesota ? Twin Cities. Funding Requests for Fundamental Research are authorized by an NSF approved solicitation, NSF 10-601. The solicitation invites I/UCRCs to submit proposals for support of industry-defined fundamental research.
Migration away from Hard Disk Drives (HDD) to solid state drives (SDDs) composed of NAND flash memory is viewed as essential to meeting access speed, energy, reliability, and mobility requirements of emerging data storage systems. The proposed effort seeks to investigate architectures for the integration of NAND flash and Phase Change Memory (PCM) ? a new technology that is highly scalable with improved performance (I/O, erase, write endurance). The proposed research plans to address design issues peculiar to the integration of these two technologies with the intent of developing new architectures, algorithms and analytic tools with which to arrive at innovative solutions that will enable use of flash-based memory devices in high-performance and low-power computer systems and data intensive applications.
NAND-PCM architectures have the potential to enhance the performance of high-end as wella s data-intensive computing while addressing increasing the increasing need to reduce energy consumption. Outcomes of the proposed research have the potential for major impact in the US computing and data systems sector. The proposed effort has garnered support from one center member, Los Alamos Nation Lab, for execution of the proposed work. This industry sector will further benefit via dissemination of the results through the Center membership and the extension of the centers project portfolio into this area. Opportunities are planned for graduate and undergraduate students including those in underrepresented groups.
The major goals of this project are to explore and understand the fundamental properities of Phase Change Memory (PCM) and Solid State Drives (SSD) and to further benefit from these new storage technologies by designing efficient ways of using them. PCM is an emerging type of Non-Volatile memory (NV-RAM) which has the potential of replacing DRAM as the main memory in current computers. SSD is flash memory based and its performance on both read and write can be 0 times fast than the current disk drives. The Pirincipal Investigators (PIs) and their team have evaluated the reliability issue of PCM. In order to accomplish this goal, a prototype system is developed using PCM chips donoted by Micro Technology. The prototype system enabled them to continuously read and write data into PCM chips. Its relability is then evaluated. The team has also developed new buffering schemes based on the special properties of SSD to improve the performance of SSD. Since SSD is essential a printed circuit board with flash memory and control unit, it is easier to include more computing power on this type of storage devices. The team has developed detailed simulation to explore such possibilities. They have investigated ways of using more computing power on a storage device to reduce the data traffic to be delivered from storage devices to computers. By this type of design, many big data probelem can be solved more efficiently. They have demonstrated that challenging problems like facial recognition and big graphs can be benefited from such new technologies.
