USENIX Courses

The Koala Benchmarks for the Shell - Characterization and Implications

Explore KOALA, a comprehensive benchmark suite for Unix/Linux shell performance research featuring diverse real-world programs, extensive analysis, and reproducible tooling infrastructure.

• YouTube
• 13 minutes
• Self-Paced
• Free Video

Opening Up Kernel-Bypass TCP Stacks

Explore comparative analysis of six kernel-bypass TCP stacks, examining their performance across diverse workloads from bulk data transfers to low-latency messaging.

• YouTube
• 17 minutes
• Self-Paced
• Free Video

Burst Computing - Quick, Sudden, Massively Parallel Processing on Serverless Resources

Explore a novel serverless solution for burst-parallel jobs with guaranteed simultaneity, optimized resource allocation, and enhanced locality for faster processing.

• YouTube
• 16 minutes
• Self-Paced
• Free Video

μEFI - A Microkernel-Style UEFI with Isolation and Transparency

Discover μEFI, a microkernel-inspired isolation framework that enhances UEFI firmware security by sandboxing modules transparently with minimal 1.91% boot overhead.

• YouTube
• 15 minutes
• Self-Paced
• Free Video

MARC - Motion-Aware Rate Control for Mobile E-commerce Cloud Rendering

Discover MARC, a motion-aware rate control framework optimizing mobile e-commerce cloud rendering by reducing session freeze rates 71% and boosting user interaction time 20%.

• YouTube
• 15 minutes
• Self-Paced
• Free Video

Accelerating Distributed Graph Learning by Using Collaborative In-Network Multicast and Aggregation

Discover how SwitchGNN accelerates distributed graph neural network training by up to 74% through innovative in-network multicast and aggregation techniques using programmable switches.

• YouTube
• 16 minutes
• Self-Paced
• Free Video

Z-LFS - A Zoned Namespace-tailored Log-structured File System for Commodity Small-zone ZNS SSDs

Discover Z-LFS, a novel log-structured file system optimized for zoned namespace SSDs, achieving up to 33.44× performance improvements through innovative metadata handling and zone management.

• YouTube
• 16 minutes
• Self-Paced
• Free Video

SNARY - A High-Performance and Generic SmartNIC-Accelerated Retrieval System

Discover SNARY, a SmartNIC-accelerated retrieval system for large-scale recommendation systems, featuring exact and fuzzy search with HBM storage and LSH-based algorithms for optimal performance.

• YouTube
• 13 minutes
• Self-Paced
• Free Video

PPipe - Efficient Video Analytics Serving on Heterogeneous GPU Clusters via Pool-Based Pipeline Parallelism

Discover how PPipe leverages pipeline parallelism to optimize video analytics on heterogeneous GPU clusters, achieving 32-75% higher throughput via pool-based resource management.

• YouTube
• 17 minutes
• Self-Paced
• Free Video

KVCache Cache in the Wild - Characterizing and Optimizing KVCache Cache at a Large Cloud Provider

Explore KVCache optimization for large language model serving through systematic workload characterization and workload-aware eviction policies at scale.

• YouTube
• 16 minutes
• Self-Paced
• Free Video

HyCache - Hybrid Caching for Accelerating DNN Input Preprocessing Pipelines

Discover HyCache, a hybrid caching runtime that accelerates DNN input preprocessing pipelines by optimally caching data across memory and storage, achieving up to 10.1× speedup improvements.

• YouTube
• 14 minutes
• Self-Paced
• Free Video

GPREEMPT - GPU Preemptive Scheduling Made General and Efficient

Discover GPREEMPT, a novel GPU preemption mechanism that achieves low-latency scheduling for diverse workloads while maintaining broad applicability to non-idempotent tasks.

• YouTube
• 12 minutes
• Self-Paced
• Free Video

PipeThreader - Software-Defined Pipelining for Efficient DNN Execution

Discover PipeThreader, a DNN compiler that shifts scheduling from hardware to software, enabling efficient computation pipelining on modern GPUs with specialized units like TensorCores.

• YouTube
• 15 minutes
• Self-Paced
• Free Video

DRack - A CXL-Disaggregated Rack Architecture to Boost Inter-Rack Communication

Discover DRack, a CXL-based rack architecture that disaggregates NICs and memory to accelerate inter-rack communication by 37.3% for data-intensive applications.

• YouTube
• 17 minutes
• Self-Paced
• Free Video

DSA-2LM - A CPU-Free Tiered Memory Architecture with Intel DSA

Discover how Intel DSA enables CPU-free tiered memory architecture, achieving up to 30% performance improvement over existing systems through adaptive page migration and optimized data movement.

• YouTube
• 17 minutes
• Self-Paced
• Free Video