Introduction to REPACSS: A Beginner’s Guide

About this page

This document introduces the foundational steps for logging in, navigating storage environments, submitting jobs, and utilizing available software on the REPACSS cluster.

Welcome to REPACSS — REmotely-managed Power Aware Computing Systems and Services at Texas Tech University. This guide is designed for new users, graduate researchers, including educators and others who are beginning their journey in high-performance computing (HPC).

---

Getting ACCESS Allocation

REPACSS is integrated with the ACCESS system. All users (TTU or external) must request compute resources through ACCESS before they can run jobs on REPACSS.

Quick Setup Steps

1. Register — Create an ACCESS account at New User Registration using your university identity.

2. Create a Project — Choose a project type (e.g., "Explore" or "Discover") at Project Types. Each type provides ACCESS credits that you can redeem for REPACSS CPU/GPU hours.

3. Exchange Credits for REPACSS — In your project's Credits + Resources tab, search for "REPACSS" and request Texas Tech REPACSS CPU (1 credit/core-hour) or GPU (100 credits/GPU hour). Complete the justification and resource questions, then submit for approval.

For detailed instructions with screenshots, see the ACCESS Account Guide.

Accessing the System

Before accessing REPACSS resources, users must be connected to TTUnet or TTUnet VPN.

TTUnet VPN Usage Cases

On Campus:
- You must be connected to TTUnet, either via wired Ethernet or the TTUnet Wi-Fi network.
- Other networks on campus (such as TTUguest or EduRoam) are not supported for direct access.
Off Campus:
- If you are connecting from any other network, including TTUguest, EduRoam, or external internet connections, you must use the TTU GlobalProtect Virtual Private Network (VPN).
- Instructions are available on the VPN Setup Guide.
Authentication: All system access requires secure login via SSH or authorized web-based interfaces.

Note: Users located within the Computer Science Department building may experience restricted access when using TTUnet Wi-Fi. If you encounter connectivity issues, connect via wired Ethernet or enable the VPN to ensure uninterrupted access.

SSH Login

To initiate a session, open a terminal (PowerShell on Windows, Terminal on macOS or Linux, or tools like MobaXterm) and run:

ssh <your_username>@repacss.ttu.edu

During first-time access, the system may prompt you to verify the server’s RSA key fingerprint. Confirm by typing yes. You will then be required to enter your password.

Note

Login nodes are reserved for light activities such as file management and job preparation. Computational jobs must be executed on compute nodes.

---

System Overview

Login Node

Login node is intended for lightweight tasks only. Users should use it to:

• Edit and manage files
• Install user-level software or modules
• Compile code (if lightweight)
• Submit SLURM job scripts

Warning

Do not run compute-intensive applications or parallel jobs on the login node.

Compute Nodes

All computational jobs should be submitted to the compute nodes via SLURM. The system includes:

• 110 CPU worker nodes for general-purpose parallel and serial computing
• 8 GPU worker nodes for accelerated workloads (e.g., deep learning, GPU-based simulations)
• 1 GPU build node for compiling and testing GPU applications in a controlled environment

Any workload requiring high-performance, extended runtime, or parallel execution should be run on these compute nodes.

System Specifications Summary Table

Node Type	Total Nodes	CPU Model	CPUs/Node	Cores/Node	Memory/Node	Storage/Node	GPUs/Node	GPU Model
CPU Nodes	110	AMD EPYC 9754	2	256	1.5 TB DDR5	1.92 TB NVMe	-	-
GPU Nodes	8	Intel Xeon Gold 6448Y	2	64	512 GB	1.92 TB SSD	4	NVIDIA H100 NVL (94 GB HBM)
Login Nodes	3	AMD EPYC 9254	2	48	256 GB	1.92 TB NVMe	-	-
Storage Nodes	9	Intel Xeon Gold (varied)	2	8–32	512 GB–1 TB	25.6–583.68 TB	-	-

---

Storage System

REPACSS offers multiple storage environments optimized for different use cases:

Storage Type	Location	Environment Variable
Home	/mnt/GROUPID/home/USERID	$HOME
Scratch	/mnt/GROUPID/scratch/USERID	$SCRATCH
Work	/mnt/GROUPID/work/USERID	$WORK

Checking Quotas

REPACSS storage space usage is organized by your assigned group.

After logging into REPACSS, run the following command:

$ df -h /mnt/$(id -gn)

Example output:

Filesystem              Size  Used Avail Use% Mounted on
10.102.95.220:/REPACSS  9.1T  162G  9.0T   2% /mnt/REPACSS

---

Software Access and Modules

REPACSS provides software access via the environment module system. This system allows users to load and unload software packages as needed.

Common Module Commands

module avail            # List available software modules
module load gcc         # Load a specific module
module list             # View loaded modules
module unload gcc       # Unload a module

Users should include required module commands at the beginning of their job scripts so the software is available when your job runs.

Tip

Not sure what's available? Run module avail first to see the full list of software on REPACSS.

Load GCC module

For example, to load gcc in a job script:

    module load gcc

For additional details, refer to the Module System documentation.

---

Job Submission with C Program

This example demonstrates the procedure for compiling and executing a basic C program using a SLURM batch script.

Create Your Source Code

Create a file called hello.c:

#include <stdio.h>

int main(){
    printf("Hello from my SLURM job.\n");
    return 0;
}

Create Your Batch Script (run_hello.sh)

#!/bin/bash
#SBATCH --job-name=hello_job        # Name of the job
#SBATCH --output=hello_output.out   # File to save standard output
#SBATCH --error=hello_error.err     # File to save error messages
#SBATCH --time=00:05:00             # Max runtime (HH:MM:SS)
#SBATCH --nodes=1                   # Number of nodes
#SBATCH --ntasks=1                  # Number of tasks
#SBATCH --cpus-per-task=1           # Number of CPUs per task

module load gcc          # Load the GCC compiler

gcc hello.c -o hello     # Compile the C program

./hello                  # Run the compiled program

To determine your resource needs, refer to the Determine Resource Needs documentation.

Submit Your Job

To submit the batch script to the SLURM workload manager, execute the following command:

sbatch run_hello.sh

Monitor Your Job

squeue --me

Check Your Output

Once the job completes, view the result:

cat hello_output.out

You should see:

Hello from my SLURM job.

How to monitor the job status?

To monitor the status of the submitted job, use:

squeue --me

For additonal details, refer to the Job Examples documentation.

---

Support and Resources

If issues arise or assistance is required, users are encouraged to:

• Refer to this user guide
• Consult with their research advisor or lab administrator
• Visit the Support Page

Tip

When seeking help, include specific error messages and a description of the attempted job to expedite troubleshooting.