Unlocking New Pose in HPC - Containerization, Cloud, and GPU-based Workloads

Overview

Explore containerization, cloud, and GPU-based workloads in HPC. Learn about Kubernetes for resource management, GPU virtualization, custom scheduling, and monitoring for efficient AI development and deployment.

Syllabus

• Introduction to HPC and AI Workloads
Overview of HPC and its relevance to AI
Introduction to GPU-based workloads
The role of cloud computing in HPC
• Fundamentals of Containerization
Basics of containers and their benefits in HPC
Container orchestration tools overview
• Deep Dive into Kubernetes
Kubernetes architecture and components
Setting up a Kubernetes cluster for HPC
Deploying and managing applications with Kubernetes
Resource management techniques in Kubernetes
• GPU Virtualization and Utilization
Understanding GPU architecture and capabilities
Tools and techniques for GPU virtualization
Best practices for maximizing GPU utilization in HPC
• Custom Scheduling in HPC
Introduction to custom schedulers in Kubernetes
Designing custom scheduling algorithms for optimized performance
Hands-on exercises with Kubernetes scheduling
• Monitoring and Performance Tuning
Monitoring tools and techniques for Kubernetes
Analyzing performance bottlenecks in GPU-based workloads
Techniques for effective performance tuning
• Cloud Integration with HPC
Exploring cloud service providers and their offerings for HPC
Design patterns for hybrid cloud HPC environments
Case studies of successful cloud-enabled HPC deployments
• Case Studies and Real-World Applications
Analyzing real-world AI applications using HPC
Lessons learned from deploying AI workloads in HPC clusters
• Future Trends and Innovations in HPC
Emerging technologies and their potential impact
Research directions in AI and HPC integration
• Course Conclusion
Recap of key concepts learned
Final project: designing a scalable AI workload deployment strategy using Kubernetes and cloud resources.

Subjects

Conference Talks