What You Need to Know Before
You Start
Starts 8 June 2025 16:08
Ends 8 June 2025
00
days
00
hours
00
minutes
00
seconds
Legion - Automatically Pushing the Envelope of Multi-GPU System for Billion-Scale GNN Training
Innovative system for accelerating billion-scale GNN training using multi-GPU caching, hierarchical partitioning, and automatic management, significantly outperforming existing solutions.
USENIX
via YouTube
USENIX
2544 Courses
20 minutes
Optional upgrade avallable
Not Specified
Progress at your own speed
Conference Talk
Optional upgrade avallable
Overview
Innovative system for accelerating billion-scale GNN training using multi-GPU caching, hierarchical partitioning, and automatic management, significantly outperforming existing solutions.
Syllabus
- Introduction to Legion
- Fundamentals of Graph Neural Networks (GNNs)
- Multi-GPU System Architecture
- Advanced Techniques in GNN Training
- Automatic Management in GNN Systems
- Legion System Design
- Comparative Analysis
- Practical Implementation
- Case Studies and Real-World Applications
- Future Directions in Multi-GPU GNN Training
- Course Wrap-up and Evaluation
Overview of Multi-GPU Systems
Challenges in Billion-Scale GNN Training
Basics of GNN Architecture
Applications of GNNs in Various Domains
GPU Basics and Parallel Processing
Multi-GPU System Integration and Communication
Hierarchical Partitioning Strategies
Multi-GPU Caching Mechanisms
Data Sharding and Distribution
Resource Allocation and Load Balancing
Dynamic Scaling and Optimization
Core Principles of Legion
Architectural Components and Innovations
Performance Metrics and Benchmarks
Comparison with Existing GNN Training Solutions
Setting up a Multi-GPU Environment
Hands-On with Legion: Installation and Configuration
Running a GNN Training Session with Legion
Successful Deployments of Legion
Insights and Lessons Learned
Emerging Trends and Technologies
Research Opportunities and Challenges
Summary and Key Takeaways
Final Assessment and Feedback
Subjects
Conference Talks
