CSR: Small: Software-defied HDDs: A System-centric Design Framework to Minimize Data Storage Cost for Data Centers

点击翻译按钮获取中文摘要

The project aims to develop storage system design techniques that can reduce the cost of hard disk drives (HDDs) for data centers. Based upon the observation that one could reduce the HDD cost by relaxing its storage reliability, the project will develop approaches that can accommodate low-cost but very unreliable HDDs in data centers without sacrificing the overall performance.The project will cohesively exploit the unique characteristics of data centers and HDDs to develop cross-layer design techniques that can most effectively embrace very unreliable HDDs in data centers. In particular, the project will develop cross-layer design techniques to achieve: (a) hybrid local/distributed sector repair, (b) iterative cross-layer decoding-based sector repair, and (c) configurable per-sector protection. The project will further investigate the integration of these design techniques and explore the involved design trade-offs. A data center HDD simulator will be implemented to support the evaluations and design space exploration, and will be made publicly available to the research community.The project could have significant impact on improving the cost effectiveness of mass data storage in data centers. The success of the project could lead to establishing a larger-scale inter-disciplinary research area covering data center storage systems, file system, coding theory, HDD firmware and read channel, and HDD manufacturing. The research training to both undergraduate and graduate students will address the concerns of lacking researchers and engineers with multi-disciplinary experience in computer industries. The main results will be research papers and a data center HDD simulator. The simulator source code and the corresponding documentation will be maintained on GitHub, where anyone can obtain and examine the most recent versions. Research papers will be available on the project web site at https://www.ecse.rpi.edu/homepages/tzhang/research.html. All the results will be kept publicly available for at least five years beyond the completion of the project.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.