CAREER: Towards Full-Stack Crash Consistency

存储系统是现代计算机的重要组成部分。每天产生的海量数据(例如金融交易、科学计算)使得存储系统中所有硬件和软件层的健壮性变得越来越重要。当面临意外的崩溃事件(例如，停电、操作系统死机)时，健壮的系统必须优雅地处理崩溃并保持崩溃的一致性，由于系统的复杂性，这在实践中是非常困难的。该项目从整体角度分析不同存储系统的崩溃一致性。该方案综合考虑了硬件特性、软件层之间的交互(如设备驱动、文件系统、键值存储)和系统执行的生命周期(如正常执行路径和恢复代码路径)以确保完整性，并利用虚拟化将干扰降至最低以实现高保真。该项目希望提高作为现代社会基本构建块的存储系统的健壮性。对于停机一分钟可能导致数百万美元损失的金融机构来说，这种健壮性的提高是有价值的；对于任何一条数据都可能促进人类知识的科学探索，其好处是不可估量的。全栈方法预计将促进计算机体系结构、操作系统、数据库、软件工程、灾难应对等方面的协同研究，并将使包括硬件制造商、软件开发人员和系统管理员在内的广泛从业者受益。此外，通过与教育和创业领域的多位专家合作，该项目将促进课程开发和行业合作，促进开源和可重复性，并扩大参与范围。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Storage systems are an essential component of modern computers. The enormous data generated every day (e.g., financial transactions, scientific computations) make the robustness of all hardware and software layers in storage systems increasingly important. When facing unexpected crash events (e.g., power outage, operating system panic), a robust system must handle the crashes gracefully and maintain crash consistency, which is extremely difficult to achieve in practice due to the system complexity. This project takes a holistic view to analyze the crash consistency of diverse storage systems. It takes into account the hardware characteristics, the interactions between software layers (e.g., device drivers, file systems, key-value stores), and the lifecycle of system execution (e.g., normal execution paths and recovery code paths) to ensure the completeness, and leverages virtualization to minimize the disturbance for high fidelity.This project hopes to advance the robustness of storage systems that serve as a fundamental building block of modern society. For financial institutions where one minute downtime may lead to millions of dollars lost, such improved robustness is valuable; for scientific exploration where any piece of data may advance human knowledge, the benefit is immeasurable. The full-stack approach is expected to stimulate synergistic research in computer architecture, operating systems, databases, software engineering, disaster response, etc., and will benefit a wide range of practitioners including hardware manufacturers, software developers, and system administrators. Moreover, by collaborating with multiple experts in education and entrepreneurship, the project will spur curriculum development and industry collaboration, promote open source and reproducibility, and broaden participation.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.