Compute architecture for accelerated AI and scientific workloads

Bring CPU, GPU, scheduler, container, storage, network and application layers into one coherent environment.

Workload Architecture

Choose the operating model that matches the workload

GPU compute, containers, notebooks, dataset services and orchestration for training and inference.

HPC

CPU/GPU nodes, MPI, Slurm/PBS, shared filesystems and scientific application stacks.

HYB

CPU science and GPU AI partitions with shared operational services and data infrastructure.

Cluster Stack

ComputeCPU nodes, GPU nodes, workstations and service nodes

ControlSlurm, PBS, Kubernetes and supporting services

NetworkManagement, data and high-speed fabric paths

StorageLocal NVMe, NFS/NAS and scalable shared storage

SoftwareMPI, CUDA, Python, containers and scientific applications

Unified stack

UsersResearchers • Engineers • Data Scientists

AccessLogin Nodes • Portals • Notebooks

SchedulingSlurm • PBS • Kubernetes

ExecutionCPU • GPU • MPI • Containers

DataShared Storage • NVMe • Datasets

Use the guided wizard to capture workload, compute, networking, storage and scheduler requirements.