SeaWulf Docs

SeaWulf Documentation

Filter by category:

No documentation matches your search criteria.

Try a different search term or filter.

Getting Started

Quick Start Guide

Essential information to get you started with SeaWulf quickly.

Cluster Overview

A high-level description of the SeaWulf cluster and its capabilities.

Glossary of Terms

Definitions of common HPC and SeaWulf-specific terminology.

HPC Cheat Sheet

A quick reference guide for common commands and operations.

Cluster Access and Accounts

Requesting Accounts & Project Numbers

How to request access to SeaWulf and get project allocations.

DUO Authentication

Setting up and using two-factor authentication for secure access.

Connecting to the Cluster

How to connect to SeaWulf using SSH, MobaXterm, Open OnDemand and other remote access tools.

Passwordless SSH

Set up SSH keys for secure, password-free login.

X11 Forwarding

How to run graphical applications on SeaWulf and display them locally.

Cluster Architecture & Resources

Architecture Overview

The overall design and capabilities of the SeaWulf cluster.

Queues Table

Detailed specifications for every queue (partition), including cores, memory, GPUs, and limits.

Queue Selection Guide

Tips, strategies, and a decision tree to help you pick the most efficient queue for your job.

Login Nodes

Details about the login nodes and their appropriate use.

28-core Intel Haswell Nodes

Specifications and best use cases for Haswell compute nodes.

40-core Intel Skylake Nodes

Specifications and best use cases for Skylake compute nodes.

96-core AMD EPYC Milan Nodes

Specifications and best use cases for AMD Milan compute nodes.

Shared Nodes

Information about shared compute resources and how to use them.

GPU Nodes

Details on GPU-accelerated compute nodes and their capabilities.

96-core HBM Sapphire Rapids Nodes

Information about high-bandwidth memory nodes and their use cases.

Job Scheduling & Resource Management

SLURM Overview

Introduction to the SLURM workload manager used on SeaWulf.

SBATCH Script Layout

How to structure job submission scripts for optimal results.

Interactive Jobs

Running interactive sessions on compute nodes.

Example Jobs

Sample job scripts for various types: MPI, OpenMP, GPU, and multi-node jobs.

Job Dependencies and Directives

Creating job workflows and using advanced SLURM directives.

Canceling and Modifying Jobs

How to manage jobs after they've been submitted.

Monitoring Jobs

Tools and commands for tracking job status and performance.

Storage and File Transfer

File Storage Layout

Overview of the storage systems available on SeaWulf.

User Storage

Information about home and scratch directories.

Project Spaces

Shared storage for research groups and collaborative projects.

File Transfer with rsync, scp, sftp

Command-line tools for transferring data to and from SeaWulf.

Globus File Transfer

Using Globus for reliable, high-performance data transfers.

CIFS Access

Mounting SeaWulf storage on your local computer (RPX and General).

Software Management

Using Modules

How to load and use software installed on the cluster.

Managing Your Own Software

Best practices for installing and managing your own software.

Conda Environments

Creating and using Conda environments for reproducible research.

Pip vs Conda Tips

Choosing between package managers and avoiding common pitfalls.

Installing from Source

Guidelines for compiling and installing software from source code.

Usage Policies and Best Practices

General Usage Guidelines

Policies and expectations for cluster usage.

Login Node Etiquette

Appropriate use of login nodes and resource limitations.

Node Utilization

Best practices for efficient use of compute resources.

Fairshare and Job Priority Tips

Understanding job scheduling priorities and maximizing throughput.

When to Ask for Help

How to troubleshoot issues and get assistance from support staff.

Open OnDemand

Overview of Open OnDemand

Introduction to the web-based portal for accessing HPC resources.

Using Jupyter

Running Jupyter notebooks on SeaWulf through the Open OnDemand web interface.

Using RStudio

Using RStudio Server for R analysis through the Open OnDemand web interface.

Using Matlab

Running Matlab through the Open OnDemand web interface.

Using Code Server

Running VS Code in your browser on SeaWulf compute resources.