USENIX Courses

Synchronizing Clocks by Leveraging Local Clock Properties

Innovative clock synchronization system leveraging local clock properties to improve accuracy and failure tolerance, reducing drift by up to 2000× for high-performance applications requiring sub-microsecond precision.

• YouTube
• 17 minutes
• Self-Paced
• Conference Talk

YuZu - Neural-Enhanced Volumetric Video Streaming

Explore neural-enhanced volumetric video streaming with YuZu, boosting quality and efficiency through 3D super resolution, QoE modeling, and adaptive resource management for immersive 6DoF experiences.

• YouTube
• 17 minutes
• Self-Paced
• Conference Talk

Spectrum - High-bandwidth Anonymous Broadcast

High-bandwidth anonymous file broadcasting system optimizing for few broadcasters and many subscribers. Offers improved throughput, malicious client prevention, and broadcaster deanonymization protection.

• YouTube
• 16 minutes
• Self-Paced
• Conference Talk

Performance Interfaces for Network Functions

Explores performance interfaces for network functions, introducing PIX tool to extract and analyze performance data, aiding developers and operators in optimizing network systems.

• YouTube
• 17 minutes
• Self-Paced
• Conference Talk

Donar - Anonymous VoIP over Tor

Explore DONAR, a system enabling anonymous VoIP over Tor with high quality-of-experience through innovative traffic management and performance optimization techniques.

• YouTube
• 15 minutes
• Self-Paced
• Conference Talk

Cetus - Releasing P4 Programmers from the Chore of Trial and Error Compiling

Automated system for converting uncompilable P4 programs into compilable versions, addressing fitting issues in programmable switches and significantly reducing development time.

• YouTube
• 16 minutes
• Self-Paced
• Conference Talk

Efficient Scheduling Policies for Microsecond-Scale Tasks

Explores efficient scheduling policies for microsecond-scale tasks in datacenters, comparing load-balancing and core allocation strategies to optimize latency and resource utilization.

• YouTube
• 15 minutes
• Self-Paced
• Conference Talk

Ekya - Continuous Learning of Video Analytics Models on Edge Compute Servers

Continuous learning for edge video analytics models, addressing data drift and balancing inference-retraining tradeoffs. Ekya optimizes resource allocation for improved accuracy with fewer GPU resources.

• YouTube
• 17 minutes
• Self-Paced
• Conference Talk

Decentralized Cloud Wide-Area Network Traffic Engineering with BLASTSHIELD

Decentralized WAN traffic engineering system slicing networks into fault domains, maximizing throughput without central coordination while reducing traffic loss from controller failures.

• YouTube
• 19 minutes
• Self-Paced
• Conference Talk

SCALE - Automatically Finding RFC Compliance Bugs in DNS Nameservers

Automated tool for finding RFC compliance bugs in DNS nameservers. Generates test cases to uncover implementation differences and vulnerabilities, improving DNS software reliability and security.

• YouTube
• 16 minutes
• Self-Paced
• Conference Talk

MLaaS in the Wild - Workload Analysis and Scheduling in Large-Scale Heterogeneous GPU Clusters

Analyze workload challenges in large-scale heterogeneous GPU clusters for ML-as-a-Service, addressing issues like utilization, queueing delays, and resource imbalance.

• YouTube
• 15 minutes
• Self-Paced
• Conference Talk

Modular Switch Programming Under Resource Constraints

Elastic data structures for optimal resource utilization in programmable networks. P4All extends P4, enabling modular switch programming and efficient resource allocation across various network applications.

• YouTube
• 14 minutes
• Self-Paced
• Conference Talk

NetVRM - Virtual Register Memory for Programmable Networks

NetVRM: A system enabling dynamic register memory sharing between applications on programmable networks, improving resource utilization and application performance through virtual memory abstraction and utility-based allocation.

• YouTube
• 16 minutes
• Self-Paced
• Conference Talk

SketchLib - Enabling Efficient Sketch-based Monitoring on Programmable Switches

Explore efficient sketch-based network monitoring on programmable switches with SketchLib. Learn optimization techniques to reduce hardware resource footprint while maintaining measurement accuracy.

• YouTube
• 15 minutes
• Self-Paced
• Conference Talk

Isolation Mechanisms for High-Speed Packet-Processing Pipelines

Explore isolation mechanisms for high-speed packet processing pipelines, focusing on Menshen's approach to enforcing isolation between modules in programmable data planes.

• YouTube
• 15 minutes
• Self-Paced
• Conference Talk