[Home]  [Publications]  [Misc] 

I joined this department in 2007. Before that, I have worked several years in Shanghai(China) after I got my master degree from Huazhong Univ. of Science and Technology and my bachelor degree from Harbin Institute of Technology . Now I am advised by Prof. Emery Berger and I am a member of PLASMA group.

• Deterministic multithreaded programming
• False sharing detection and mitigation
• Tolerating concurrency errors
• Resource management of operating systems

Dthreads: Deterministic multithreaded library (SOSP 2011)

Dthreads is an efficient deterministic multithreading system for unmodified C/C++ applications by replacing the existing pthreads library.
Dthreads enforces determinism even in the face of data races. Dthreads works by exploding multithreaded applications into multiple processes, with private, copy-on-write mappings to shared memory. It uses standard virtual memory protection to track writes, and deterministically orders updates by each thread. By separating updates from different threads, Dthreads has the additional benefit of eliminating false sharing. Experimental results show that Dthreads substantially outperforms a state-of-the-art deterministic runtime system, and for a majority of the benchmarks evaluated here, matches and occasionally exceeds the performance of pthreads.

Sheriff: False sharing detection and mitigation (OOPSLA/SPLASH 2011)

We provide two tools to handle false sharing problems based on the Sheriff framework:
Sheriff-Detect can find instances of false sharing by comparing updates within the same cache lines by different threads, and uses sampling to rank them by performance impact. Sheriff-Detect is precise (no false positives), runs with low overhead (on average 20%), and is accurate, pinpointing the exact objects involved in false sharing.
Sheriff-Protect mitigates false sharing problems by adaptively isolating shared updates from different threads into separate physical addresses, effectively eliminating most of the performance impact of false sharing. Sheriff-Protect can be used in the following cases: it is infeasible to fix false sharing problems when source is unavailable, or undesirable when padding objects would unacceptably increase memory consumption or further worsen runtime performance.

Grace: Safe Multithreaded Programming for C/C++(OOPSLA 2009)

Grace is a runtime system for multithreaded programs, which provides good scalability and performance while eliminating a range of concurrency errors. To exploit available computing resources (multiple CPUs or cores), Grace employs a combination of speculative thread execution, supported by a novel virtual-memory based software transactional memory system, together with a sequential commit protocol.

Redline: First Class Support of Interactivity in Commodity Operating Systems (OSDI 2008)

Redline provides support for interactive applications in commodity operating systems. Redline works with unaltered applications and uses standard APIs. Users specify resource requirements for programs by lightweight specifications, Redline can enforce these resource allocations on different components of operating systems, including CPU scheduler, memory manager and IO scheduler.

Reduce memory fragmentation:

In this project, we tackled the memory fragmentation problems in Linux operating system. By combining "memory allocation according to usage types" and "page migration", we can greatly reduce the memory fragmentation.