USENIX Courses

AC-Key - Adaptive Caching for LSM-based Key-Value Stores

Adaptive caching technique for LSM-based key-value stores, improving read performance by managing multiple cache components and adjusting sizes based on workload demands.

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

The Future of the Past - Challenges in Archival Storage

Explores challenges and advancements in long-term digital data preservation, discussing media technologies, data integrity, and semantic interpretation for future generations.

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

A Decentralized Blockchain with High Throughput and Fast Confirmation

Explore a novel blockchain system with high throughput and fast confirmation, featuring an optimistic consensus protocol and adaptive weight mechanism to enhance scalability and security.

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

Reexamining Direct Cache Access to Optimize I-O Intensive Applications for Multi-hundred-gigabit Networks

Optimizing direct cache access for high-speed networks: Analyzes DDIO implementation, demonstrates latency reduction for 100 Gbps networks, and highlights challenges at 200 Gbps, offering insights for system optimization.

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

BatchCrypt - Efficient Homomorphic Encryption for Cross-Silo Federated Learning

Efficient homomorphic encryption system for cross-silo federated learning, reducing encryption and communication overhead while maintaining privacy and accuracy in collaborative machine learning.

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

MatrixKV - Reducing Write Stalls and Write Amplification in LSM-tree Based KV Stores

Innovative LSM-tree based key-value store design using NVM to reduce write stalls and amplification. Introduces matrix container, column compaction, and cross-row hint search for improved performance in multi-tier storage systems.

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

BASTION - A Security Enforcement Network Stack for Container Networks

Innovative security enforcement network stack for containers, enhancing visibility and isolation while improving performance. Mitigates attacks and optimizes inter-container communication.

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

SweynTooth - Unleashing Mayhem over Bluetooth Low Energy

Explores vulnerabilities in Bluetooth Low Energy protocol implementations, presenting a systematic fuzzing framework that discovered 11 new vulnerabilities across various devices and IoT products.

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

coIOMMU - A Virtual IOMMU with Cooperative DMA Buffer Tracking for Efficient Memory Management in Direct I/O

Explore efficient memory management in direct I/O virtualization through coIOMMU, a novel virtual IOMMU architecture with cooperative DMA buffer tracking for improved performance and resource utilization.

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

Twizzler - A Data-Centric OS for Non-Volatile Memory

Explore Twizzler, an innovative OS redesign for non-volatile memory, offering efficient data persistence, simplified programming, and improved performance over traditional systems.

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

Can Applications Recover from fsync Failures?

Análisis de cómo sistemas de archivos y aplicaciones modernas reaccionan ante fallos de fsync, revelando vulnerabilidades críticas en la durabilidad de datos y sugiriendo mejoras de diseño.

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

Serverless in the Wild - Characterizing and Optimizing the Serverless Workload

Explore serverless computing workloads, challenges, and optimization strategies in cloud environments. Gain insights into Azure Functions' production data and learn about innovative resource management policies.

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

PinK - High-speed In-storage Key-value Store with Bounded Tails

Explore high-speed in-storage key-value systems with PinK, an innovative LSM-tree-based design that significantly improves performance and reduces latency in resource-constrained environments.

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

Offload Annotations - Bringing Heterogeneous Computing to Existing Libraries and Workloads

Explore offload annotations for efficient GPU computing in existing libraries, enabling seamless integration and performance gains without extensive code modifications.

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

HetPipe - Enabling Large DNN Training on Heterogeneous GPU Clusters through Integration of Pipelined Model Parallelism and Data Parallelism

Explores HetPipe, a system integrating pipelined model parallelism and data parallelism for training large DNNs on heterogeneous GPU clusters, improving performance and enabling use of less powerful GPUs.

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