OSDI (Operating Systems Design and Implementation) Talks & Presentations

USENIX

Virtual Consensus in Delos

Explores virtualizing consensus in replicated systems, enabling dynamic protocol changes without downtime. Introduces VirtualLog and Loglets for simplified, diverse, and composable consensus implementations, showcasing benefits in the Delos database.

• YouTube
• 18 minutes
• On-Demand
• Conference Talk

USENIX

Beyond Malloc Efficiency to Fleet Efficiency - A Hugepage-Aware Memory Allocator

Explore TEMERAIRE, a hugepage-aware memory allocator enhancing TCMALLOC to reduce CPU overheads, improve application performance, and optimize fleet efficiency in warehouse-scale computing.

• YouTube
• 14 minutes
• On-Demand
• Conference Talk

USENIX

Oort - Efficient Federated Learning via Guided Participant Selection

Efficient federated learning system that improves model training and testing by intelligently selecting participants based on data utility and device capabilities, enhancing accuracy and performance.

• YouTube
• 16 minutes
• On-Demand
• Conference Talk

USENIX

Rearchitecting Linux Storage Stack for µs Latency and High Throughput

Explore a novel Linux kernel storage stack architecture achieving microsecond-scale latency and high throughput, adapting network switch techniques for improved performance without application modifications.

• YouTube
• 15 minutes
• On-Demand
• Conference Talk

USENIX

The nanoPU - A Nanosecond Network Stack for Datacenters

Innovative NIC-CPU co-design accelerating datacenter applications with μs-scale RPCs. Features fast network-to-CPU path, hardware-supported transport and load balancing, achieving 69ns wire-to-wire response time and improved performance.

• YouTube
• 15 minutes
• On-Demand
• Conference Talk

USENIX

Pollux - Co-adaptive Cluster Scheduling for Goodput-Optimized Deep Learning

Innovative deep learning cluster scheduler optimizing resource allocation and job performance through adaptive co-optimization, resulting in reduced completion times and improved fairness among competing jobs.

• YouTube
• 14 minutes
• On-Demand
• Conference Talk

USENIX

Distributed Deep Graph Learning at Scale

Explores a novel system for scaling Graph Neural Network training to large real-world graphs, introducing pipelined push-pull parallelism for faster and more efficient distributed processing.

• YouTube
• 14 minutes
• On-Demand
• Conference Talk

USENIX

Twine - A Unified Cluster Management System for Shared Infrastructure

Explore Facebook's Twine, a unified cluster management system for shared infrastructure, scaling to manage one million machines across data centers with innovative approaches to workload customization and resource allocation.

• YouTube
• 19 minutes
• On-Demand
• Conference Talk

USENIX

Theseus - An Experiment in Operating System Structure and State Management

Redesigning OS modularity using Rust, Theseus reduces inter-component state dependencies, enabling live evolution and fault recovery for core components beyond existing systems.

• YouTube
• 20 minutes
• On-Demand
• Conference Talk

USENIX

RedLeaf - Isolation and Communication in a Safe Operating System

Explore RedLeaf, a Rust-based OS leveraging language safety for fine-grained isolation. Learn about innovative approaches to fault isolation, driver recovery, and performance in operating system design.

• YouTube
• 22 minutes
• On-Demand
• Conference Talk

USENIX

Ansor - Generating High-Performance Tensor Programs for Deep Learning

Ansor: A framework for generating high-performance tensor programs in deep learning, improving execution speed across various hardware platforms through innovative search strategies and optimization techniques.

• YouTube
• 20 minutes
• On-Demand
• Conference Talk

USENIX

Serving DNNs like Clockwork - Performance Predictability from the Bottom Up

Explore a novel approach to machine learning inference systems that achieves predictable end-to-end performance, supporting thousands of models while meeting strict latency targets and request-level SLOs.

• YouTube
• 20 minutes
• On-Demand
• Conference Talk

USENIX

Heterogeneity-Aware Cluster Scheduling Policies for Deep Learning Workloads

Explores heterogeneity-aware scheduling for deep learning workloads, introducing Gavel to optimize resource allocation in clusters with diverse accelerators, improving efficiency and performance.

• YouTube
• 20 minutes
• On-Demand
• Conference Talk

USENIX

Scaling Distributed Machine Learning with the Parameter Server

Explore a framework for distributed machine learning using parameter servers, enabling scalable processing of petabyte-scale data with billions of examples and parameters across various algorithms.

• YouTube
• 24 minutes
• On-Demand
• Conference Talk

USENIX

Assise - Performance and Availability via NVM Colocation in a Distributed File System

Explores a novel distributed file system leveraging persistent memory modules for enhanced IO performance, rapid failover, and strong consistency, outperforming traditional remote storage models.

• YouTube
• 21 minutes
• On-Demand
• Conference Talk