Overview of NRP and the Nautilus Cluster

The National Research Platform (NRP) is a unique partnership of over 50 institutions, led by researchers at UC San Diego, University of Nebraska-Lincoln, and the Massachusetts Green High Performance Computing Center. Supported by the National Science Foundation (NSF), Department of Energy (DOE), and Department of Defense (DoD) among others, the NRP is a community-owned research and education platform. Its mission is to connect researchers and educators, fostering collaboration, accelerating innovation, and sharing resources. The NRP provides access to cutting-edge technologies in AI, high-performance computing, data storage, and networking, available free of charge to non-profit research and education institutions. As of early 2024, the NRP connects over 400 nodes across 70+ locations on 3 continents, serving over 5,000 users.

The NRP Nautilus Cluster is a key component of this national infrastructure. Nautilus is a HyperCluster specifically designed for running containerized Big Data and AI applications. It leverages Kubernetes (K8S) for managing applications and Rook for Ceph data services automation. Nautilus features a diverse range of computational resources, including CPUs, GPUs (from consumer-grade to high-end AI accelerators like NVIDIA A100s and H100s), FPGAs, and a federated national-scale CDN.

Data Policy: The NRP Nautilus Cluster, like all NRP resources, currently has no storage suitable for HIPAA, PID, FISMA, FERPA, or protected data of any kind. Users are not permitted to store such data on NRP machines.

Getting Started with Nautilus

For UCR-specific assistance or initial inquiries, please contact:

• Email: research-computing@ucr.edu
• Slack: UCR Research Computing Slack

For general NRP support and documentation, refer to the following resources:

• NRP Documentation Home: The main portal for all Nautilus and NRP user guides.
• Getting Access to Nautilus: Steps to obtain an account and access the Nautilus cluster.
• Nautilus Quick Start Guide: A beginner’s guide to quickly start using Nautilus.
• Nautilus Cluster Policies: Important operational policies for using the cluster.
• NRP Acceptable Use Policy (AUP): Overall AUP for the National Research Platform.
• NRP/Nautilus FAQ: Answers to frequently asked questions about Nautilus.
• General NRP FAQs: Broader questions about the National Research Platform.
• Glossary of Terms: Definitions of common terms used in the NRP ecosystem.
• Asking for Support (Matrix/Chat): How to get help from the NRP support team.

Using the Nautilus Cluster

This section provides links to essential documentation for leveraging the capabilities of the Nautilus Cluster.

Core Concepts & Tutorials

• General Guide to Using Nautilus: An overview of best practices and how to effectively use the platform.
• Introduction to Docker and Containers: Fundamental concepts for working with containerized applications.
• Basic Kubernetes on Nautilus: Learn the basics of deploying and managing applications with Kubernetes.
• Working with Container Images: How to find, use, and manage container images, including Scientific Images.
• Scheduling Pods and Jobs: Understanding how to schedule your workloads effectively.
• Debugging Applications on Nautilus: Tips and techniques for troubleshooting.

Running Computational Tasks

• Running GPU Pods: How to request and utilize GPU resources.
• Running Batch Jobs: For non-interactive, compute-intensive tasks. See also: Optimizing for High I/O Jobs.
• Running CPU-only Jobs: For workloads that do not require GPUs.
• Long-running Pods and Services: Guidelines for services that need to run continuously.
• Monitoring your Pods and Jobs: Keeping track of your resource usage and job status.
• For more advanced topics, see the full Running on Nautilus Documentation.

Storage on Nautilus

Nautilus offers a variety of storage solutions to meet different research needs.

• Storage Options Overview: An introduction to the available storage types.
• Ceph FileSystem (FS): Provides home directories and project spaces. Accessed via PersistentVolumeClaim in Kubernetes. (Details)
• Ceph S3 Object Storage: Scalable object storage compatible with the S3 API. (Details)
• CVMFS (CernVM File System): Used for distributing software and data efficiently. (Details)
• Local Scratch Storage: Fast, temporary storage available on worker nodes. Data is not persistent. (Details)
• Data Management:
  • Moving Data In and Out: Includes information on tools like kubectl cp, rclone, and Globus Connect.
  • Purging Data: Policies and procedures for data lifecycle management.
  • Managing S3 Tokens/Credentials.
• For other storage solutions like Linstor or Nextcloud, refer to the main Nautilus Storage Documentation.

AI and Machine Learning

Nautilus is well-equipped for a wide range of AI and Machine Learning workloads.

• Overview of AI/ML on NRP: General information and resources.
• Using GPUs for AI/ML: Essential for training and inference.
• NRP Managed LLM Service: Access pre-configured Large Language Models.
• Running LLMs in JupyterHub: Integrate LLMs into your Jupyter environment.
• Using Qualcomm Cloud AI 100 Cards: For specialized AI acceleration.

Jupyter and Interactive Computing

• Nautilus JupyterHub Service: Easy access to JupyterLab instances.
• Deploying a Custom ML/Jupyter Pod: For tailored Jupyter environments with specific packages.
• Deploying Your Own JupyterHub: For groups or classrooms needing a shared JupyterHub instance.
• Using Coder for an IDE-like Environment: Access a VS Code-like interface in your browser. (Deploying Coder)

Development and Version Control

• Building Container Images in GitLab CI/CD: Automate your image building process.
• Using Private Container Registries/Repositories.
• Integrating GitLab and Kubernetes: Streamline your development workflow.

Specialized Hardware and Networking

• Using FPGAs:
  • Vivado and Vitis for Xilinx FPGAs
  • ESnet SmartNIC Development
• Network Experiments with FABRIC Integration: For advanced networking research.

Additional Resources & Community

• Deployed Services on NRP: Explore other tools and services available on the platform (many require separate registration).
• NRP Cluster Dashboards & Visualizations:
  • Main Dashboard
  • Cluster Nodes Map
  • Busy Cluster Map (Traceroute View)
• NRP Blog/News: Stay updated on the latest developments and events.
• Educational Use / Classroom Resources: Information for educators wishing to use NRP in their curriculum.

The NRP Nautilus Cluster, as part of the broader National Research Platform, offers a rich set of resources and a collaborative environment to advance computational research and education.