Cluster blueprint catalog

This page describes Google's catalog of cluster blueprints, which are reusable YAML configuration files for deploying specific, opinionated solutions with the Cluster Toolkit. Each blueprint defines a specific environment, including the scheduler, storage, machine types, and base operating system.

You can filter the catalog by your required scheduler, storage, machine types, and base operating system. Each entry highlights featured applications and Google Cloud resources that the blueprint leverages.

A cluster blueprint is a YAML file that defines a reusable configuration and describes the specific environment that you want to deploy using Cluster Toolkit. A cluster blueprint implements and documents a specific opinionated solution. For more information about cluster blueprints, see Cluster blueprint.

To search the catalog, you can filter by a combination of your required scheduler, storage, machine types, and base operating system. For each cluster blueprint, featured applications and Google Cloud resources that can be leveraged by using the blueprint are also highlighted.

Cluster blueprint Scheduler Storage Machine types Base operating system Featured software and compute resources Collection
a4-highgpu-8g Slurm Filestore, Cloud Storage, Google Cloud Managed Lustre A4 ubuntu-accelerator-2204-amd64-with-nvidia-570
  • NVIDIA ConnectX-7
  • Google Virtual NIC (gVNIC)
  • GPUs
  • Virtual Private Cloud
 Core
gke-a3-ultragpu GKE Filestore, Cloud Storage A3 Ultra Container-Optimized OS
  • NVIDIA ConnectX-7
  • Google Virtual NIC (gVNIC)
  • GPUs
  • Virtual Private Cloud
 Core
gke-a3-highgpu GKE None A3 High Container-Optimized OS
  • Jobset
  • Kueue
 Core
gke-a3-megagpu GKE None A3 Mega Container-Optimized OS
  • Jobset
  • Kueue
 Core
gke-managed-hyperdisk GKE Filestore, Cloud Storage C3 Container-Optimized OS Core
gke-a4 GKE A4 Container-Optimized OS
  • Jobset
  • Kueue
 Core
a3-ultragpu-8g Slurm Filestore, Cloud Storage, Google Cloud Managed Lustre A3 Ultra ubuntu-accelerator-2204-amd64-with-nvidia-570
  • NVIDIA ConnectX-7
  • Google Virtual NIC (gVNIC)
  • GPUs
  • Virtual Private Cloud
 Core
a3-highgpu-8g Slurm Filestore, Cloud Storage A3 High ubuntu-2004-lts*
  • Google Virtual NIC (gVNIC)
  • GPUs
  • Virtual Private Cloud
 Core
a3-megagpu-8g Slurm Filestore, Cloud Storage A3 Mega debian-12*
  • Google Virtual NIC (gVNIC)
  • GPUs
  • Virtual Private Cloud
 Core
h4d-vm None Filestore H4D hpc-rocky-linux-8
  • Google Virtual NIC (gVNIC)
  • Cloud RDMA
  • Local SSDs
 Core
hpc-slurm-h4d Slurm Filestore N2, H4D hpc-rocky-linux-8
  • Google Virtual NIC (gVNIC)
  • Cloud RDMA
  • Local SSDs
 Core
hpc-slurm Slurm Filestore N2, C2, H3 hpc-rocky-linux-8* Core
hpc-slurm6-tpu-maxtext Slurm None TPUv4, N2, C2 slurm-gcp-6-4-tf-2.14.0*
  • TPUs
  • Maxtext
 Community, Experimental
ml-slurm Slurm Filestore, Cloud Storage N2, A2, G2 debian-11*
  • GPUs
  • NCCL
  • PyTorch
  • TensorFlow
 Core
serverless-batch Batch Filestore N2 rocky-linux-8-optimized-gcp Core
serverless-batch-mpi Batch Filestore C2  hpc-rocky-linux-8
  • Spack
  • Weather Research and Forecasting (WRF)
 Core
tutorial-starccm None Filestore C2  hpc-centos-7
  • Star CCM+
 Community, Experimental
tutorial-fluent None Filestore C2  hpc-rocky-linux-8
  • Ansys Fluent
 Community, Experimental
image-builder Slurm Filestore C2, N2 hpc-rocky-linux-8*
  • Packer
 Core
hpc-build-slurm-image Slurm None N2D hpc-rocky-linux-8*
  • Packer
  • Building Slurm
 Community
hpc-amd-slurm Slurm Filestore C2D hpc-rocky-linux-8*
  • AMD
  • AMD Optimizing C/C++ and Fortran Compilers (AOCC)
  • OpenFoam
  • Spack
 Community
hpc-slurm-local-ssd Slurm Filestore N1 hpc-rocky-linux-8*
  • Local SSDs
 Community, Experimental
hpc-slurm-ubuntu2004 Slurm Filestore N2, C2  ubuntu-2004-lts Community
hpc-slurm-gromacs Slurm Filestore C2  hpc-rocky-linux-8*
  • GROMACS
  • Spack
 Community, Experimental
hpc-slurm-hcls Slurm Cloud Storage, Filestore A2, C2  hpc-rocky-linux-8*
  • GPUs
  • GROMACS on GPUs
  • Spack
  • VDI with VMD
 Community
htc-htcondor HTCondor None C2  hpc-rocky-linux-8 Community, Experimental
ml-gke GKE None N2, A2 Container-Optimized OS
  • GPUs
 Community, Experimental
hpc-gke GKE None N2, C2  Container-Optimized OS Community, Experimental
storage-gke GKE Filestore, ephemeral storage N2D Container-Optimized OS
  • Local SSDs
  • CSI Drivers
  • Persistent Volumes
 Community, Experimental
client-google-cloud-storage None Cloud Storage E2 hpc-rocky-linux-8 Community, Experimental
hpc-slurm-sharedvpc Slurm Filestore N2, C2  hpc-rocky-linux-8*
  • Shared VPC configuration
 Community

*This OS image is an optimized Schedmd's Slurm image that is based on the specified operating system.

†This OS image has DAOS drivers pre-installed and needs to be built following instructions in the DAOS project.

What's next