ABR: CSR: Medium: Collaborative Research: FTFS: A Read/Write Optimized Fractal Tree File System

点击翻译按钮获取中文摘要

Computer applications rely on file systems to read, write, and access user data safely and efficiently. Unfortunately, conventional file system designs make some applications fast at the expense of others; similarly, most file systems are designed to perform well on one particular type of storage device, such as a hard disk drive (HDD) or flash-based solid-state drive (SSD). Currently, a wide variety of end-users do not realize the full performance potential of their storage hardware, unless they are able to select a file system that is well-matched to their specific workload and hardware. Prior work on the BetrFS file system prototype demonstrated that, with better on-disk data structures, it is possible to build a single file system that roughly matches or exceeds the performance of the best file system for any workload on an HDD. This project will extend these performance gains, which are enabled by a tight, cross-disciplinary collaboration between systems and theory researchers, to a wider range of storage hardware, ranging from complex, archival devices (such as shingled magnetic recording drives), to very fast flash devices (such as non-volatile memory express devices). This project will also investigate a more efficient approach to directory-level cloning, a useful building block for both server workloads and personal computing systems. By making a large swath of applications run faster across a range of storage hardware, this project can have significant real-world benefits. Storage hardware is becoming increasingly varied, and this project can help future-proof the performance of any application, from commerce to health-care to scientific computing to education. This project involves an important component to transition these ideas from research to practice, including maturing the open-source BetrFS prototype, as well as continuing outreach and tutorials for practitioners. This project will continue training both undergraduate and graduate students to design, analyze, build, and measure algorithmically sound computer systems that are unlike anything they have seen previously.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.