OSG workshop for students
Date: Wednesday, May 31 from 1:00-3:00.
Max Attendees: 35
Title: Scale out your computing on the OSPool
Summary: The OSPool is a national computing resource perfect for running computational work that can be scaled out as tens, hundreds, and thousands of computational tasks. This training will provide researchers with a hands-on introduction to using the OSPool, including example use cases and how to scale them out.
Prerequisites: Some very basic shell knowledge is helpful but not strictly required.
Kubernetes workshops 1
This is a series of three workshops. Attending all is not required.
Date: Wednesday, May 31 from 10:15 – 12:00
Title: Introduction to Kubernetes
In this workshop, researchers from the University of Missouri will discuss building reproducible and portable containers for software using Docker and deploying them to Kubernetes. Each step of the deployment from local development machine to Kubernetes cluster will be covered, including building and pushing custom containers to image registries, building Kubernetes pod and job YAML files, and deploying pods to the cluster. Finally, this hands-on tutorial will show how to use a web-enabled development environment to build and deploy containers with Kubernetes using nothing but a web-browser
- Introduction to Docker
- Containerization of Software
- Building and pushing custom containers
- Introduction to Kubernetes
- Kubernetes Architecture and Concepts
- Interacting with Kubernetes using KubeCTL
- Building Deployment YAML Files for Kubernetes
- Introduction to the NRP Nautilus HyperCluster
- Deploying Pods and Jobs in Kubernetes using Nautilus
- Utilizing Git-Pod to Deploy Containers from the Web
kubernetes workshop 2
Date: Wednesday May 31, 1:00-2:30
Title: Migrating Research to NRP
In this workshop, researchers from the University of Missouri will discuss migrating deep learning research using Kubernetes on the NRP Nautilus HyperCluster. Nautilus contains over 1300 NVIDIA GPUs and 20,000 CPU cores that can be leveraged to scale deep learning research beyond a single GPU-enabled computer. This workshop will cover how to build and deploy containers onto Nautilus for the training and testing of PyTorch deep learning models for multiple computer vision tasks, including classification, object detection and localization, and semantic segmentation. Finally, this workshop will discuss the ability to automate the deployment of training jobs to Kubernetes to vastly scale training workflows and train several models in parallel.
- Building Optimized Containers for Deep Learning with PyTorch
- Using Common Frameworks: Detectron2 and MMDetection
- NRP Nautilus Hyper Cluster
- Kubernetes Architecture and Key Concepts
- Introduction to NRP Nautilus HyperCluster
- Migrating Data to NRP with S3
- Deploying GPU Jobs to Nautilus for Training Computer Vision Models
- Automating GPU Jobs on Nautilus using Bash and Python
kubernetes workshop 3
Date: Thursday, June 1, 2:30 – 4:00
Title: Advanced Workshop: Troubleshooting Kubernetes and Generating Metrics
Date: Wednesday, May 31 from 1:00-4:00
Title: Interactive Cybersecurity Readiness Exercise
Summary: This workshop will allow participants to evaluate their cybersecurity posture by working through real world scenarios. Participants will be networking, information sharing and discussing best practices with each other.