Taming Wild Concurrent Programs

The reality of multi-core hardware has made concurrent programs pervasive. Unfortunately, writing correct concurrent programs is difficult. As programmers are used to sequential thinking, concurrent programs can easily run wild. Additionally, due to the non-deterministic nature, concurrency bugs are notoriously hard to expose during testing and to reproduce for diagnosis.

In this talk, I will describe our recent research effort in combating wild concurrent programs. More specifically, I will talk about our findings in (1) hardware and system support as well as using data mining and machine learning/NLP techniques to automatically infer programmers' synchronization intent and detect concurrency bugs; (2) systematic approaches to identify and cover corner-case interleavings and expose more concurrency bugs during in-house testing; and (3) real world concurrency bug characteristics that can shed some lights on future research on concurrency bug detection, testing and programming language design.

Bio:Yuanyuan Zhou is an associate professor in the Department of Computer Science at Univ of Illinois at Urbana Champaign. Prior to UIUC, she worked at NEC Research Institute as a scientist after completing her Ph.D at Princeton in 2000. Her research interests span the areas of operating systems, architecture, storage systems and software reliability. She was the recipient for the Alfred Sloan Fellowship 2007, UIUC Gear Faculty Award 2006, NSF Career-2004 award, the CRA-W Anita-Borg Early Career Award 2005, the DOE Early Career Principle Investigator Award 2005, the IBM Faculty Award 2004 & 2005, and the IBM SUR-2003 award. She has 3 papers selected into the IEEE Micro Special Issue on Top Picks from Architecture Conferences and one best paper in SOSP 2005. She was also selected into the "Incomplete List of Teachers Ranked as Excellent by Their Students” in 2003 and 2006 at UIUC.