Ordered Systems Lab at U-M

Research Interests

Our research spans broadly across operating systems, distributed systems, cloud computing, mobile systems, and ML infrastructure, while specializing in reliability, fault tolerance, and performance. Our work combines systems building with deep insights to address real-world challenges facing modern systems and enable ORDER. Our research innovations cover:

Foundation: Design formal reasoning techniques to ensure strong correctness guarantees in complex systems.
Abstraction: Introduce new abstractions and interfaces to address fundamental gaps while avoiding ad-hoc designs.
Analysis: Develop automated program analyses, data-driven methods, and ML techniques to better reason about system behavior.
Runtime: Build robust runtime mechanisms to observe, mitgate, and recover from assorted issues and enable self-adapting systems.

ORDER := {Observable, Reliable, Defensible, Efficient, Responsive}

News

Dec 2025 Yuzhuo successfully defended his PhD thesis titled "Operating System Support for Reliable Software" and will join Google after graduation. Congratulations, Dr. Jing!
Jul 2025 TrainVerify is accepted to SOSP '25! TrainVerify uses equivalence-based verification to provide strong correctness guarantess for the parallelization logic of distributed LLM training.
Jul 2025 Phoenix is accepted to SOSP '25! Phoenix provides OS-level support for optimistic recovery and partial state preservation for high-availability software.
Jul 2025 Atropos is accepted to SOSP '25! Atropos is an application overload control framework that uses targeted cancellation to maintain tight SLOs.
Mar 2025 TrainCheck is accepted to OSDI '25! TrainCheck automatically infers invariants tailored for DL training and uses these invariants to proactively detect silent training errors.