HPC Class

Introduction
Requesting Access to Class Partitions
Hardware Overview
Accessing HPC Class Partitions
Launching Jobs in Class Partitions
Class Storage

Introduction

HPC Class partitions are Slurm partitions on ISU HPC's Nova cluster that are dedicated to education and class use.

Requesting Access to Class Partitions

HPC Class partitions are available for classes. Instructors on record can request access to Class partitions for themselves and the students on their class lists via this form.

Return to Table of Contents

Hardware Overview

Number of Nodes	Processors per Node	Cores per Node	Memory per Node	Local $TMPDIR Disk	GPU	Interconnect
28	Two 2.6 GHz 32-core Intel 8358 CPUs	64	500 GB Available	1.6 TB NVMe	N/A	100 Gb/s IB
3	Two 2.65 GHz 24-core AMD Epyc 7413 CPUs	48	500 GB Available	960 GB NVMe	Eight Nvidia A100 80GB GPUs	100 Gb/s IB

Return to Table of Contents

Accessing HPC Class Partitions on Nova

The HPC Class partitions are Slurm partitions on ISU HPC's Nova Cluster. Follow the directions to login to the Nova cluster here: Nova Access and Login Guide

Partition Name	Max Time (Hrs:Min:Sec)	Nodes in Partition (Variable)
class-short	00:15:00	22
class-long	12:00:00	8
class-gpu	06:00:00	2
class-gpu-short	00:15:00	1

Return to Table of Contents

Launching Jobs in Class Partitions

To launch a job in Class partitions, specify a class partition with -p. To specify the class account to run Slurm jobs with, use the -A option. For example, a user in the class ABCD495 in Fall of 2022 with an associated Slurm account of f2022.ABCD.495.1 could run the salloc command:

salloc -p class-short -N 1 -n 4 -t 15 -A f2022.ABCD.495.1

Class instructors and TAs will need to specify account class-faculty instead:

salloc -p class-short -N 1 -n 4 -t 15 -A class-faculty

An sbatch script to request the same allocation would be as follows:

#!/bin/bash
# Copy/paste this job script into a text file and submit with the command:
#    sbatch thefilename
# job standard output will go to the file slurm-%j.out (where %j is the job ID)
#SBATCH --time=00:15:00   # walltime limit (HH:MM:SS)
#SBATCH --nodes=1   # number of nodes
#SBATCH --ntasks-per-node=4   # 4 processor core(s) per node
#SBATCH --partition=class-short    # class node(s)
#SBATCH --account=f2022.ABCD.495.1    #account to use
# LOAD MODULES, INSERT CODE, AND RUN YOUR PROGRAMS HERE

To view current Class partitions and their status, a user could run the sinfo command and search for 'class':

sinfo | grep class

Note: Users with multiple classes that utilize the cluster or users encountering "Invalid account or account/partition combination specified" errors should specify their relevant class account.

For more information on Slurm commands, see the Managing jobs using Slurm Workload Manager guide. For sample job scripts to use with the sbatch command, see the Slurm script generator for Nova.

Return to Table of Contents

Class Storage

There are three class-specific storage locations on Nova:

/work/class-faculty - Location for instructors to store course documents and files.
/work/classtmp - Temporary storage location for job data. Available to all students and instructors. The files will be deleted at semester end. Files that are desired to be retained should be moved from /work/classtmp before then.
/work/class-old-home - Read-only export of the old home directories of the defunct HPC-Class standalone cluster which was replaced in August 2022 by HPC Class partitions on Nova. If you were a recent user and had files in your home directory on the old Class cluster, you can recover them from this location.

Return to Table of Contents