Jetstream2 GPU

Jetstream2 GPU is a cloud-based computing resource that provides on-demand virtual machines with GPU acceleration for research and education. Unlike traditional HPC systems, users launch and manage their own virtual machines rather than submitting jobs to a shared scheduler.

GPU-enabled instances are available through Jetstream2 GPU flavors and may provide NVIDIA A100, L40S, or H100 GPU resources depending on the selected flavor and allocation access. Jetstream2 GPU is suitable for workloads such as machine learning, AI model training, GPU-accelerated simulations, visualization, and interactive or long-running research software environments.

Jetstream2 is built on OpenStack and provides flexible infrastructure through interfaces such as Exosphere, Horizon, CACAO, and the OpenStack command line. Users should choose VM flavors according to workload needs and SU usage, and should review Jetstream2 guidance before resizing full-card GPU instances.

GET AN ACCOUNT ON JETSTREAM2 GPU

JetStream2 ExoSphere

Jobs

Jetstream2 GPU does not use a centralized job scheduler. Workloads are executed directly on GPU-enabled virtual machines selected by flavor.

Users run workloads interactively, through scripts, containers, notebooks, services, or other software installed inside the VM environment. For example:

python train_model.py

If batch scheduling is required, users may deploy and manage their own scheduler or orchestration layer, such as Slurm or Kubernetes, within their virtual infrastructure.

Resources are consumed based on the selected GPU VM flavor and runtime. Users should shelve instances when they are not in use to conserve SUs.

Queue specifications

Name Purpose CPUs GPUs RAM Jobs
30 days
Wait Time
30-day trend
Wall Time
30-day trend
Indiana Jetstream2 GPU GPU-enabled virtual machine flavors; not a Slurm partition Varies by flavor; 8–128 vCPUs NVIDIA A100, L40S, and H100 GPU flavors; partial, single-GPU, and multi-GPU options depend on flavor and allocation access Varies by flavor; 30–960 GB VM RAM

Software

No software usage data is currently reported for Jetstream2 GPU in XDMoD.

SEE ALL SOFTWARE AVAILABLE ON JETSTREAM2 GPU


Storage

File System

Directory Path Quota Purge Backup Notes

External Storage

Storage Filesystems

Jetstream2 does not use traditional shared HPC file systems. Storage is managed through cloud-style storage resources attached to or accessed from virtual machines.

Instance Storage

  • Root or local disk associated with a VM instance.
  • Useful for operating system files, installed software, temporary data, and runtime files.
  • Data on an instance root disk can be lost when the instance is deleted, depending on how the instance was created.

Persistent Volumes

  • Block storage volumes that can be attached to VM instances.
  • Useful for datasets, results, database storage, larger working directories, and custom boot disk sizes.
  • Volumes are generally attached to a single active instance at a time.
  • Volumes can be detached and reattached to another instance when needed.

Manila Shares

  • Shared file storage that can be mounted by multiple VM instances.
  • Useful for collaborative datasets, class/workshop material, and shared input/output data for distributed workloads.
  • Implemented using OpenStack Manila.

Object Store

  • S3-compatible object storage based on OpenStack Swift.
  • Useful for public data hosting, web application assets, S3-compatible tools, and HTTP-accessible data.
  • Available through Horizon and the OpenStack command line interface.

Jetstream2 does not provide a general backup service for user data. Users should keep offsite copies of important data.


File Transfer

Jetstream2 does not use dedicated data transfer nodes. File transfers are performed directly to and from Jetstream2 virtual machine instances or through web-based and managed transfer tools.

Common transfer methods include the Exosphere web shell/web desktop file-transfer interface, SCP/SFTP to a VM with a public IP address, and Globus for larger or managed transfers. The best method depends on file size, whether the VM has a public IP address, and whether the data is being transferred from a local machine, another endpoint, or another digital resource.

Supported Methods Data Transfer Node URL
GLOBUS | RECOMMENDED https://www.globus.org/data-transfer
EXOSPHERE WEB SHELL/WEB DESKTOP https://docs.jetstream-cloud.org/ui/exo/exo-filetransfer/
SCP
SFTP