OSDI 2025

Meta Info

Homepage: https://www.usenix.org/conference/osdi25

Acceptance Rate

14.6% (= 48 / 327)

Papers

Large Language Models (LLMs)

• LLM Training

  • WLB-LLM: Workload-Balanced 4D Parallelism for Large Language Model Training

    • UCSD

• LLM Inference

  • Fast and Live Model Auto Scaling with O(1) Host Caching

    • SJTU IPADS

Deep Learning Compilation

• KPerfIR: Towards a Open and Compiler-centric Ecosystem for GPU Kernel Performance Tooling on Modern AI Workloads

  • UCSD

• Mirage: A Multi-Level Superoptimizer for Tensor Programs [arXiv] [Code]

  • CMU

GPU Sharing

• Preemptive Scheduling for Diverse XPUs using Multi-level Hardware Model

  • SJTU IPADS

GPU Communication

• Enabling Efficient GPU Communication over Multiple NICs with FuseLink

  • HKUST iSING Lab

Resource Allocation

• Decouple and Decompose: Scaling Resource Allocation through a Different Lens

  • Harvard

Memory Translation

• EMT: An OS Framework for New Memory Translation Architectures

  • UIUC

Vector Search

• Quake: Adaptive Indexing for Vector Search

  • UW-Madison

File Systems

• Fast and Synchronous Crash Consistency with Metadata Write-Once File System

  • HIT-SZ

Databases

• Tigon: A Distributed Database for a CXL Pod

  • UT-Austin

Replicated State Machines (RSMs)

• Picsou: Enabling Efficient Cross-Consensus Communication [arXiv]

  • UC Berkeley