MoatLab

We build computer systems that bridge operating systems and computer architecture. Our work spans memory and storage systems, closing the gap between fast-moving hardware and the software that has to run on it.

We currently work in four directions to advance systems support for emerging hardware and data-intensive workloads:

1. Memory tiering and pooling. Treating memory-access criticality as a first-class design axis, we make CXL memory usable at datacenter scale (Pond, SoarAlto, PACT).
2. Performance modeling. Deriving heterogeneous-memory behavior from first principles, we make performance a design-time property rather than a deployment surprise (Melody, CAMP).
3. Tools. Building emulators, profiling frameworks, and benchmarks as research primitives, we empower the broader systems community to ask new questions across the stack (FEMU, Cylon, WARP).
4. Programmability. Co-designing the OS-hardware interface, we build programming models for computational and disaggregated hardware (LeapIO, IODA, RAIZN).