USENIX Courses

Is Big Data Performance Reproducible in Modern Cloud Networks?

Explores network variability in cloud computing, its impact on big data workloads, and provides guidelines for more repeatable and predictable performance experiments in cloud environments.

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

Learning Relaxed Belady for Content Distribution Network Caching

Innovative machine learning approach for CDN caching, approximating Belady MIN algorithm. Reduces WAN traffic, outperforms state-of-the-art methods, with modest overhead for real-world deployment.

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

Programmable Calendar Queues for High-Speed Packet Scheduling

Explores programmable calendar queues for dynamic packet scheduling, enabling priority changes based on network time. Demonstrates implementation and benefits through case studies in delay guarantees, fairness, and flow prioritization.

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

Finding Network Misconfigurations by Automatic Template Inference

Automated outlier detection technique for network configurations, inferring templates to identify likely errors without explicit specifications. Applicable to complex elements like ACLs and route policies.

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

Ghostor - Toward a Secure Data Sharing System from Decentralized Trust

Explore a secure data-sharing system that protects user privacy and ensures data integrity through decentralized trust, hiding identities and detecting server-side violations.

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

Config2Spec - Mining Network Specifications from Network Configurations

Automated system for synthesizing formal network specifications from configurations, combining data plane analysis and control plane verification to enhance network reliability and security.

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

Adapting TCP for Reconfigurable Datacenter Networks

Explores TCP adaptation for modern reconfigurable datacenter networks, proposing dynamic queue resizing and explicit circuit state feedback to improve circuit utilization and performance.

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

XRD - Scalable Messaging System with Cryptographic Privacy

Scalable messaging system offering cryptographic privacy, using parallel mix networks and novel techniques to support millions of users with low latency while protecting communication metadata.

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

Meaningful Availability

Explores a new availability metric for cloud applications: windowed user-uptime. Discusses its advantages over traditional metrics in accurately reflecting user experience and distinguishing outage patterns.

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

Firecracker - Lightweight Virtualization for Serverless Applications

Explore Firecracker, a lightweight VMM for serverless applications, offering strong security and minimal overhead. Learn about its design, deployment, and impact on AWS Lambda and Fargate services.

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

Experiences with Modeling Network Topologies at Multiple Levels of Abstraction

Explores MALT, a multi-layer topology representation for network management, covering its benefits, implementation challenges, and lessons learned in creating a comprehensive topology model.

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

Expanding Across Time to Deliver Bandwidth Efficiency and Low Latency

Innovative network design for datacenters combining low-latency delivery and high-bandwidth efficiency through dynamic reconfiguration, outperforming static topologies for various workloads.

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

AmphiLight - Direct Air Water Communication with Laser Light

Innovative laser-based system for high-bandwidth bidirectional air-water communication, adapting to water dynamics and offering 3D steering capabilities for improved underwater operations and coordination.

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

Themis - Fair and Efficient GPU Cluster Scheduling

Innovative GPU cluster scheduling framework for fair and efficient allocation of resources in distributed machine learning workloads, addressing unique challenges of long-running, gang-scheduled tasks and placement sensitivity.

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

DC-Store - Eliminating Noisy Neighbor Containers Using Deterministic I/O Performance and Resource Isolation

Explore a storage framework offering deterministic I/O performance in multi-container environments, eliminating noisy neighbor issues through hardware and software innovations.

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