CAREER: Scalable and Maximal Concurrency Debugging

标题：Career：可伸缩和最大并发调试与并发相关的软件缺陷在实践中是最昂贵和最危险的缺陷之一，并已成为对安全关键系统和国家关键基础设施的可靠性和安全性的主要威胁。该项目的智力优势在于开发了新的理论和算法进步，并构建了用于调试并发软件的实用自动化工具，这将使更多的并发错误更容易检测和理解，并更安全、更快地修复。该项目的更广泛的意义和重要性是帮助生产更可靠、更安全、更经济的软件系统和基础设施，消除与并发相关的漏洞，并加强STEM软件调试教育，这是当今软件工程教育的一个关键但缺乏的方面。(1)复制：如何在长时间运行的并发程序中以最小的运行时扰动和开销复制失败？如何在运行在普通硬件上的并发程序中再现表现出放松内存行为的故障？(2)检测：如何最大限度地检测并发错误，并且没有错误警报，即使在有限的观察下，例如当丢失日志事件时？(3)理解：如何准确地识别故障的根本原因？如何有效地简化并发错误并加速其复制？(4)修复和验证：如何在不引入死锁或不必要的性能降级的情况下修复并发错误？如何有效地验证修复的正确性？

Title: CAREER: Scalable and Maximal Concurrency DebuggingConcurrency related software defects are among the most expensive and dangerous in practice, and have become a major threat to the reliability and security of safety-critical systems and the nation's critical infrastructure. The intellectual merits of this project are to develop new theoretical and algorithmic advances, and to build practical automated tools for debugging concurrent software, which will enable more concurrency bugs easier to detect and understand, and safer and faster to fix. The project's broader significance and importance are to help produce more reliable, secure, and economical software systems and infrastructure, remove concurrency related vulnerabilities, and strengthen STEM education on software debugging, which is a critical, but lacking aspect of today's software engineering education.This project will investigate solutions to four major research questions on concurrency debugging. (1) Reproduction: How to reproduce failures in long running concurrent programs with minimal runtime perturbation and overhead? How to reproduce failures in concurrent programs running on commodity hardware exhibiting relaxed-memory behaviors? (2) Detection: How to detect concurrency bugs at the maximum ability and with no false alarm, even under limited observation, e.g., when missing log events? (3) Understanding: How to accurately identify the failure's root cause? How to effectively simplify concurrency bugs and speed up their reproduction? (4) Fixing & Validation: How to fix concurrency bugs without introducing deadlocks or unnecessary performance degradation? How to effectively validate the correctness of fixes?