USENIX Courses

NetTLP - A Development Platform for PCIe Devices in Software Interacting with Hardware

Explore NetTLP, a platform for developing PCIe devices that enables software-hardware interaction. Learn about its potential impact on datacenter networks and low-latency, high-bandwidth packet delivery.

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

SP-PIFO - Approximating Push-In First-Out Behaviors Using Strict-Priority Queues

Innovative packet scheduling technique using strict-priority queues to approximate PIFO behavior, enabling programmable scheduling at line rate and scale on existing hardware.

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

TCP ≈ RDMA - CPU Efficient Remote Storage Access with i10

Innovative remote storage stack i10 achieves RDMA-like performance on commodity hardware, saturating 100Gbps links with low CPU usage while allowing unmodified applications to use the kernel's TCP/IP stack.

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

ABC - A Simple Explicit Congestion Controller for Wireless Networks

Innovative congestion control protocol for wireless networks, offering improved throughput and lower delays compared to existing solutions. Easily deployable without changes to headers or devices.

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

FPGA-Accelerated Compactions for LSM Based Key-Value Store

Explore FPGA acceleration for LSM-tree key-value stores, improving compaction speed and system throughput while reducing CPU bottlenecks and enhancing energy efficiency in short KV storage scenarios.

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

InfiniCache - Exploiting Ephemeral Serverless Functions to Build a Cost-Effective Memory Cache

Innovative serverless-based object caching system offering cost savings, high data availability, and competitive performance for storage-intensive web applications.

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

How to Copy Files

Explores efficient file cloning techniques in BetrFS, introducing "nimble clones" that optimize performance and space usage through copy-on-abundant-write policy and write-optimized key-value store.

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

Strong and Efficient Consistency with Consistency-Aware Durability

Introduces consistency-aware durability for strong consistency in distributed storage. Demonstrates cross-client monotonic reads in ORCA, a modified ZooKeeper, offering improved performance over strongly consistent systems.

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

Gandalf - An Intelligent, End-To-End Analytics Service for Safe Deployment in Large-Scale Cloud Infrastructure

Intelligent analytics service for safe software deployment in large-scale cloud infrastructure, monitoring fault signals and assessing rollout impact to prevent widespread outages.

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

CarMap - Fast 3D Feature Map Updates for Automobiles

Explores crowd-sourced 3D map collection for autonomous vehicles, presenting CarMap's lean representation and efficient updating techniques to enable near real-time uploads over cellular networks.

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

Introduction to Client Side Caching in Ceph

Explore Ceph's client-side caching evolution, focusing on RBD storage policies and upcoming write-back cache implementation for SSDs and persistent memory to enhance distributed storage performance.

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

Millions of Tiny Databases

Innovative database system for cloud storage, optimizing availability and consistency through distributed architecture. Explores Physalia's design and applications in AWS, offering insights for distributed systems.

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

Making Disk Failure Predictions SMARTer

Explore advanced machine learning techniques for predicting disk failures in large-scale data centers, improving reliability and reducing downtime in IT infrastructure.

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

Aperture - A Non-Cooperative, Client-Side Load Balancing Algorithm

Explore Twitter's Aperture algorithm for non-cooperative, client-side load balancing in large-scale RPC frameworks, addressing scalability challenges in service clusters with thousands of instances.

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

Kubernetes the Very Hard Way

Dive into complex runtime and networking challenges of large-scale Kubernetes clusters. Learn architectural improvements for scalability and performance, and explore ecosystem tools for enhanced cluster management.

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