DataScience Workbook / 06. High-Performance Computing (HPC) / 1. Introduction to HPC infrastructure / 1.3 ISU HPC

HPC Clusters at ISU

High Performance Computing (HPC) refers to a cluster of interconnected computers or servers used to perform complex computational tasks. HPC clusters can process large amounts of data in a short time and can handle large computationally intensive jobs. General information about HPC clusters available at Iowa State University can be found at the ISU HPC Website ⤴.

Components of HPC Clusters at ISU

headnode or login node - used to log in, submit jobs, and view output
data transfer node - used for data transfers
compute nodes - different types including regular, fat, GPU, etc. are used for computations
Infiniband switch - connects all nodes

Research Clusters

Condo Computing Cluster - Free access cluster
Nova Computing Cluster - Faculty can purchase nodes and storage on this cluster

Education Cluster

HPC Class - These are partitions on Nova cluster dedicated to class use
- Check the Guide to HPC Class ⤴ from ISU for more details

Storage

Large Scale Storage - Storage service for ISU researchers

Logging in

To log in to Nova/Condo Clusters, you need to be connected to ISU network on-campus or use ISU VPN ⤴ if you are off-campus. The first time you try to log in to Nova or Condo, a new Google Authenticator (GA) account is created. You will need to use GA for login. The first time you use the Secure Shell ⤴ connection to Nova/Condo, it will respond with:

"host key not found, generate hostkey?(yes/no)"

You need to answer yes to continue.

Condo vs Nova Cluster

Feature	Condo	Nova
Login node	@condo2017.its.iastate.edu	@nova.its.iastate.edu
Data Transfer node	@condodtn.its.iastate.edu	@novadtn.its.iastate.edu
Cores per node	16/20/32/40	36/48/64/96
Home directory space	5 GB	5 GB
Workload Manager	SLURM	SLURM
Partitions	Partitions List ⤴	Partitions List ⤴

Best Practices: Tips

Test your code/job on a small scale before submitting it to the cluster, if possible. This can help save time and resources as submitting a job often takes longer due to queue times and uses more resources. For example, you can use interactive nodes to find and fix errors before submitting a job.
Make sure your script is optimized for the cluster such that it uses the available resources efficiently.
Use proper documentation and version control such as Git to track changes and ease of reproducibility.
Manage your files appropriately on the cluster. Cleaning up unnecessary files and organising data/files is important.

Also check out the Project Management ⤴ guide for additional tips.

Where to find help?

FAQs ⤴ - List of Frequently asked questions such as:
Contact: hpc-help@iastate.edu

Explore hands-on tutorials

Tutorials to get you started on different HPC Clusters are available in this workbook.

Condo Computing Cluster ⤴
Nova Computing Cluster ⤴
Large Scale Storage ⤴
Software Available on HPC ⤴
Check out the SLURM Basics ⤴ and Creating SLURM Job Submission Scripts ⤴ for running your jobs on any HPC cluster.
It is useful to have some basic knowledge about command line before using HPC. Check this Unix tutorial ⤴ to learn some basics.

Guides provided by ISU

General guides

Data Transfers

For transferring data, use the File Transfers ⤴ guide. This uses Condo Data Transfer Node as an example (condodtn). Each cluster has its own data transfer node, for example, Nova cluster has novadtn.
Use Rclone ⤴ guide to sync data between HPC Clusters and Google Drive
Globus Connect ⤴ online data transfer system. Use Globus Connect Personal to copy data between a cluster and your personal computer

ISU HPC

Sharu Paul