MoatLab Research Group | MoatLab

PACT: A Criticality-First Design for Tiered Memory

Thu, 02 Apr 2026 00:00:00 +0000

*Jinshu Liu and Hanchen Xu contributed equally to this work.

Conference: ASPLOS'26 (30th ACM International Conference on Architectural Support for Programming Languages and Operating Systems)

To Appear

Performance Predictability in Heterogeneous Memory

Wed, 01 Apr 2026 00:00:00 +0000

Conference: ASPLOS'26 (31st ACM International Conference on Architectural Support for Programming Languages and Operating Systems)

To Appear

Characterizing and Emulating FDP SSDs with Warp

Sun, 01 Feb 2026 00:00:00 +0000

Conference: FAST'26 (24th USENIX Conference on File and Storage Technologies)

To Appear

Cylon: Fast and Accurate Full-System Emulation of CXL-SSDs

Sun, 01 Feb 2026 00:00:00 +0000

*Dongha Yoon and Hansen Idden are co-lead authors.

Conference: FAST'26 (24th USENIX Conference on File and Storage Technologies)

To Appear

SGDRC: Software-Defined Dynamic Resource Control for Concurrent DNN Inference on NVIDIA GPUs

Wed, 01 Jan 2025 00:00:00 +0000

Conference: PPoPP'25 (30th ACM SIGPLAN Annual Symposium on Principles and Practice of Parallel Programming)

This paper presents SGDRC, a software-defined dynamic resource control system for concurrent DNN inference on NVIDIA GPUs.

Systematic CXL Memory Characterization and Performance Analysis at Scale

Wed, 01 Jan 2025 00:00:00 +0000

Conference: ASPLOS'25 (30th ACM International Conference on Architectural Support for Programming Languages and Operating Systems)

This paper presents a systematic characterization and performance analysis of CXL memory at scale.

Tiered Memory Management Beyond Hotness

Wed, 01 Jan 2025 00:00:00 +0000

Conference: OSDI'25 (19th USENIX Symposium on Operating Systems Design and Implementation)

This paper presents novel approaches to tiered memory management that go beyond traditional hotness-based policies.

Storage

Sun, 15 Dec 2024 00:00:00 +0000

Overview

We focus on breaking through the performance barriers of modern storage systems. We explore novel storage architecture, addressing the growing demands of modern datacenter applications.

Key Research Areas

Hardware-Software Co-Design

Custom storage accelerators and programmable devices
Operating system optimizations for next-generation storage
Cross-layer optimization from applications to hardware

Impact and Applications

Our storage technologies target critical challenges in:

Cloud Infrastructure: Enabling efficient resource utilization in large-scale deployments
Data-Intensive Computing: Accelerating analytics and machine learning workloads
Enterprise Systems: Improving reliability and performance for mission-critical applications

Dissecting CXL Memory Performance at Scale: Analysis, Modeling, and Optimization

Sun, 01 Dec 2024 00:00:00 +0000

Memory Disaggregation

Sun, 01 Dec 2024 00:00:00 +0000

Operating Systems

Tue, 15 Oct 2024 00:00:00 +0000

Tuning Fast Memory Size based on Modeling of Page Migration for Tiered Memory

Tue, 01 Oct 2024 00:00:00 +0000

Preprint: arXiv:2410.00328, October 2024

This paper introduces Tuna, a system to decide fast memory size based on modeling of page migration for tiered memory systems.

Missile: Fine-Grained, Hardware-Level GPU Resource Isolation for Multi-Tenant DNN Inference

Mon, 01 Jul 2024 00:00:00 +0000

ZoneTrace: A Zone Monitoring Tool for F2FS on ZNS SSDs

Mon, 01 Jan 2024 00:00:00 +0000

Journal: ACM Transactions on Design Automation of Electronic Systems (TODAES), Volume 29, Issue 2, 2024

This paper presents ZoneTrace, a zone monitoring tool for F2FS on ZNS SSDs.

Is Garbage Collection Overhead Gone? Case study of F2FS on ZNS SSDs

Sat, 01 Jul 2023 00:00:00 +0000

Pond: The Case of CXL Memory Pooling for Cloud Datacenters

Wed, 01 Mar 2023 00:00:00 +0000

Design Tradeoffs in CXL-Based Memory Pools for Cloud Platforms

Sun, 01 Jan 2023 00:00:00 +0000

Journal: IEEE Micro, Volume 43, Issue 3, 2023

This paper analyzes design tradeoffs in CXL-based memory pools for cloud platforms.

Extending and Programming the NVMe I/O Determinism Interface for Flash Arrays

Sun, 01 Jan 2023 00:00:00 +0000

Journal: ACM Transactions on Storage (TOS), Volume 19, Issue 4, 2023

This paper extends and programs the NVMe I/O Determinism Interface for flash arrays, providing deterministic I/O performance.

Pond: CXL-Based Memory Pooling Systems for Cloud Platforms

Sun, 01 Jan 2023 00:00:00 +0000

Conference: ASPLOS'23 (28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems)

This paper presents Pond, a CXL-based memory pooling system for cloud platforms.

RAIZN: Redundant Array of Independent Zoned Namespaces

Sun, 01 Jan 2023 00:00:00 +0000

Conference: ASPLOS'23 (28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems)

This paper presents RAIZN, a redundant array system designed specifically for zoned namespaces (ZNS) SSDs, providing fault tolerance and performance optimization for modern storage systems.

First-generation memory disaggregation for cloud platforms

Tue, 01 Mar 2022 00:00:00 +0000

Fantastic SSD Internals and How to Learn and Use Them

Sat, 01 Jan 2022 00:00:00 +0000

Conference: SYSTOR'22 (15th ACM International Systems and Storage Conference)

This paper explores SSD internals and provides methods to learn and utilize them effectively.

IODA: A Host/Device Co-design for Strong Predictability Contract on Modern Flash Storage

Mon, 25 Oct 2021 00:00:00 +0000

Conference: 28th ACM Symposium on Operating Systems Principles (SOSP ‘21), October 25-28, 2021, Virtual Event, Germany

This paper presents IODA, a comprehensive framework for achieving I/O determinism in flash storage arrays. It provides programming interfaces and system support for predictable I/O performance through host/device co-design.

Evolving Storage Stack for Predictability and Efficiency

Mon, 01 Jun 2020 00:00:00 +0000

LeapIO: Efficient and Portable Virtual NVMe Storage on ARM SoCs

Sun, 01 Mar 2020 00:00:00 +0000

Fail-Slow at Scale: Evidence of Hardware Performance Faults in Large Production Systems

Thu, 01 Mar 2018 00:00:00 +0000

Fail-Slow at Scale: Evidence of Hardware Performance Faults in Large Production Systems

Thu, 01 Feb 2018 00:00:00 +0000

The CASE of FEMU: Cheap, Accurate, Scalable and Extensible Flash Emulator

Thu, 01 Feb 2018 00:00:00 +0000

MittOS: Supporting Millisecond Tail Tolerance with Fast Rejecting SLO-aware OS interface

Sun, 01 Oct 2017 00:00:00 +0000

Tiny-Tail Flash: Near-Perfect Elimination of Garbage Collection Tail Latencies in NAND SSDs

Thu, 01 Jun 2017 00:00:00 +0000

Tiny-Tail Flash: Near-Perfect Elimination of Garbage Collection Tail Latencies in NAND SSDs

Wed, 01 Feb 2017 00:00:00 +0000