USENIX Courses

Nomad: Non-Exclusive Memory Tiering via Transactional Page Migration

Innovative memory management strategy for tiered systems, improving performance under pressure through non-exclusive page retention and transactional migration.

• YouTube
• 15 minutes
• On-Demand
• Free Video

ACCL+: FPGA-Based Collective Engine for Distributed Applications

Explore ACCL+, an open-source FPGA-based collective communication library for distributed applications, offering portability, multiple protocol support, and user extensibility for enhanced performance.

• YouTube
• 15 minutes
• On-Demand
• Free Video

Fairness in Serving Large Language Models

Explore innovative scheduling algorithms for fair and efficient LLM inference services, addressing challenges of unpredictable request lengths and unique batching characteristics on parallel accelerators.

• YouTube
• 16 minutes
• On-Demand
• Free Video

Burstable Cloud Block Storage with Data Processing Units

Innovative I/O scheduling system for cloud block storage, enhancing performance and reducing latency. Utilizes data processing units and queue scaling to improve load balancing and mitigate tenant interference.

• YouTube
• 14 minutes
• On-Demand
• Free Video

Motor: Enabling Multi-Versioning for Distributed Transactions on Disaggregated Memory

Innovative system for efficient distributed transactions on disaggregated memory, improving throughput and reducing latency through multi-versioning and RDMA-based protocols.

• YouTube
• 14 minutes
• On-Demand
• Free Video

Ransom Access Memories - Achieving Practical Ransomware Protection in Cloud with DeftPunk

Innovative ransomware protection system for cloud block stores. DeftPunk uses two-layer classification, strategic snapshots, and log-structured support for efficient detection and recovery with minimal data loss.

• YouTube
• 15 minutes
• On-Demand
• Free Video

Parrot: Efficient Serving of LLM-based Applications with Semantic Variable

Discover Parrot, a system enhancing LLM-based applications with Semantic Variables. Learn how this innovation optimizes performance and enables efficient programming of AI agents and co-pilots.

• YouTube
• 13 minutes
• On-Demand
• Free Video

Microkernel Goes General: Performance and Compatibility in the HongMeng Production Microkernel

Explore HongMeng, a general-purpose microkernel addressing performance and compatibility challenges while maintaining security and reliability for smartphones and smart vehicles.

• YouTube
• 16 minutes
• On-Demand
• Free Video

Llumnix - Dynamic Scheduling for Large Language Model Serving

Innovative system for efficient LLM serving, using dynamic rescheduling to improve load balancing, resource utilization, and request prioritization, resulting in better performance and cost savings.

• YouTube
• 16 minutes
• On-Demand
• Free Video

Ladder: Enabling Efficient Low-Precision Deep Learning Computing through Hardware-aware Tensor Transformation

Innovative compiler for optimizing low-precision deep learning computations, bridging custom data types with hardware limitations to enhance performance without hardware modifications.

• YouTube
• 16 minutes
• On-Demand
• Free Video

Fast and Scalable In-network Lock Management Using Lock Fission

Innovative distributed lock service using programmable switches for improved latency and throughput, supporting millions of locks through lock fission and on-demand migration techniques.

• YouTube
• 14 minutes
• On-Demand
• Free Video

dLoRA - Dynamically Orchestrating Requests and Adapters for LoRA LLM Serving

Innovative system for efficient LoRA model serving, dynamically orchestrating requests and adapters to improve throughput and reduce latency in LLM inference.

• YouTube
• 15 minutes
• On-Demand
• Free Video

Managing Memory Tiers with CXL in Virtualized Environments

Explores hardware-managed memory tiering with CXL, introducing Intel® Flat Memory Mode and Memstrata allocator to optimize performance and reduce contention in virtualized cloud environments.

• YouTube
• 16 minutes
• On-Demand
• Free Video

Using Dynamically Layered Definite Releases for Verifying the RefFS File System

Explore RefFS, a groundbreaking concurrent file system with proven liveness and safety. Learn about dynamically layered definite releases and the MoLi framework for verifying concurrent systems.

• YouTube
• 15 minutes
• On-Demand
• Free Video

MAST: Global Scheduling of ML Training across Geo-Distributed Datacenters at Hyperscale

Intelligent global scheduling system for ML training workloads across datacenters, optimizing resource utilization and eliminating regional overload through innovative design principles.

• YouTube
• 16 minutes
• On-Demand
• Free Video