CNS Core: Small: RUI: A Holistic Approach to Taming Unaligned Writes in Flash Drives

Washington State University, Pullman WA

Investigators

Abstract

Flash based solid-state disks (SSDs) are replacing hard-disk drives in various storage systems from high-end servers in data centers to mobile computers on the edges of the cloud. Due to the architectural nature of SSDs, performance-sensitive applications may generate a large number of unaligned writes, which may lead to load imbalance and write space amplification in the constituent non-volatile memory chips. This project will investigate a new approach called WAFlash, to alleviate the side-effects of unaligned writes in SSDs. To eliminate the unaligned writes caused by suboptimal data layout, this project will investigate new strategies to store multiple valid versions of the data in flash to remove the read-modify-write operations from the critical I/O (Input/Output) path, and a page-compaction scheme to further reduce the number of partial-page writes to circumvent write amplification. This project includes four research tasks: (1) design of partial-page congestion-aware buffer replacement schemes, (2) design of flash translation layer (FTL) to support multi-version partial-pages in SSDs, (3) FTL algorithms supporting partial page compaction, and (4) system implementation and evaluation. The project will benefit a wide variety of scientific, and business intelligence applications that are mission-critical, latency-sensitive, and/or data-intensive because SSDs play an important role in the storage market. Research findings will be integrated into undergraduate-level courses as OpenSSD labs. As an RUI (Research in Undergraduate Institution) project, the investigator will provide research training for undergraduate students. Investigator will also involve high schools, delivering lectures on computer systems research, and encourage women and minority students to pursue information technology related careers. Post-publication data including code and results from the WAFlash project will be stored in a public code repository https://gitlab.encs.vancouver.wsu.edu/xuechen.zhang/waflash.git as open source. All data are stored in electronic format for at least 5 years. 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.