主讲人：Yin Wang讲座时间：３月７日（周四）上午十点地点：软件楼 102 第二会议室联系人：唐恩美 emtang@fudan.edu.cn
Abstract: Software failures are extremely costly. As programs become increasingly concurrent and distributed, defects are more likely to occur but harder to diagnose and repair. Software fault and failure detection is an active research area, but remedies remain mostly manual, expensive, and error-prone. This talk describes a method that automatically avoids software failures at runtime using Discrete Control Theory (DCT), a mature branch of control theory targeting systems with discrete state space and event-driven dynamics. Applied to concurrency bug elimination, DCT provides strong correctness and performance guarantees that follow directly from its mathematical foundations. Our approach minimizes runtime instrumentation overhead and scales to real software including OpenLDAP, BIND, and OpenJMS, automatically eliminating both known and newly detected bugs while imposing very modest runtime overhead.Bio: Yin Wang received B.S. (2000) and M.S. (2003) degrees from Shanghai Jiao Tong University. He earned his Ph.D at the University of Michigan EECS department and joined HP Labs in 2008. While a graduate student, he interned at Microsoft Shanghai, IBM Almaden Research Center, and HP Labs. Wang’s research interest is in the area of computer systems in general. His dissertation research applies control theory to concurrency bug elimination; it was featured in IEEE Computer magazine, received a CACM Research Highlights nomination from SIGPLAN, and an ACM Doctoral Dissertation Award nomination from the University of Michigan. This research work spawned many follow-up projects and led to several dedicated workshops and invited sessions. Wang continues to orchestrate several of the spinoff projects through ongoing collaborations with several universities worldwide.
