USENIX Courses

Deriving Semantic Checkers from Tests to Detect Silent Failures in Production Distributed Systems

Discover how to automatically derive semantic checkers from test code to detect silent failures in distributed systems using the T2C framework.

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

PathWeaver - A High-Throughput Multi-GPU System for Graph-Based Approximate Nearest Neighbor Search

Discover PathWeaver, a novel multi-GPU framework that accelerates graph-based approximate nearest neighbor search with 3.24× speedup through GPU-aware pipelining and optimized data selection.

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

ShieldReduce - Fine-Grained Shielded Data Reduction

Discover ShieldReduce, a secure storage system combining deduplication, delta compression, and local compression in Intel SGX for high storage savings with data confidentiality.

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

Deterministic Client - Enforcing Determinism on Untrusted Machine Code

Explore DeCl, a sandboxing system that enforces deterministic behavior on untrusted x86-64 and Arm64 machine code using Software Fault Isolation techniques for enhanced security.

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

IRHash - Efficient Multi-Language Compiler Caching by IR-Level Hashing

Discover IRHash, an innovative LLVM-based compiler caching system using IR-level hashing for multi-language support, achieving 19% build time reduction across C, C++, Fortran, and Haskell projects.

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

Mako - Speculative Distributed Transactions with Geo-Replication

Discover Mako, a high-performance transactional key-value store using speculative execution and geo-replication to achieve 8.6× higher throughput than existing systems.

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

Principles and Methodologies for Serial Performance Optimization

Discover systematic approaches to sequential task optimization through three core principles and eight actionable methodologies, plus explore SysGPT for performance suggestions.

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

Learning-Enhanced High-Throughput Pattern Matching Based on Programmable Data Plane

Discover Trochilus, a novel framework combining machine learning with programmable data planes for high-throughput network pattern matching and automated security threat detection.

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

SAVE - Software-Implemented Fault Tolerance for Model Inference against GPU Memory Bit Flips

Discover SAVE, a software system protecting ML model inference from GPU memory bit flips in safety-critical applications with minimal performance overhead.

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

WIC - Hiding Producer-Consumer Synchronization Delays with Warp-Level Interrupt-based GPU Communications

Discover WIC, a novel GPU synchronization framework that replaces polling with warp-level interrupts, improving performance by 1.13× across diverse communication patterns.

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

Stripeless Data Placement for Erasure-Coded In-Memory Storage

Discover Nos, a stripeless erasure coding scheme that eliminates stripe-based overheads in fast in-memory storage systems using XOR encoding and SBIBD structures for optimal performance.

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

SpaceExit - Enabling Efficient Adaptive Computing in Space with Early Exits

Discover SpaceExit, an adaptive computing system for satellites that optimizes Earth observation imagery processing through geospatial-contextual detection and complexity-driven scheduling.

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

DShuffle - DPU-Optimized Shuffle Framework for Large-scale Data Processing

Discover how DShuffle leverages DPUs to accelerate shuffle operations in distributed data processing, reducing CPU overhead and improving Spark performance through optimized pipelining.

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

Understanding Stragglers in Large Model Training Using What-if Analysis

Explore straggler issues in large language model training through comprehensive what-if analysis using real ByteDance cluster data and performance impact studies.

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

Understanding and Detecting Fail-Slow Hardware Failure Bugs in Cloud Systems

Discover how fail-slow hardware bugs impact cloud systems and learn about Sieve, a testing framework that detects these critical vulnerabilities in distributed systems.

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