Collaborative AI development
AI Lab
Workstations, GPU compute, shared datasets, containers, Python environments, monitoring and controlled user access.
Discuss this solution ↗REFERENCE SOLUTIONS
Architectures for AI labs,
HPC clusters and datacenters.
Reference patterns can be adapted to workload, scale, budget, performance and operational requirements.
SOLUTION OPTIONS
Reference patterns for your environment
Scientific and parallel workloads
HPC Cluster
Login and management nodes, CPU/GPU compute, schedulers, MPI, shared storage, high-speed networking and monitoring.
Discuss this solution ↗Shared platform for mixed workloads
Hybrid AI + HPC
Separate CPU and GPU workload resources while sharing suitable operational services, storage and network infrastructure.
Discuss this solution ↗Integrated compute and data layers
AI Factory / Datacenter
Plan compute, accelerators, networking, storage, Linux, orchestration, monitoring and application support as a connected environment.
Discuss this solution ↗AI LAB
A collaborative environment for AI development.
Combine workstations, GPU compute, high-speed networking, shared datasets, AI frameworks, monitoring and controlled user access.
GPU servers
CPU service nodes
High-speed networking
Shared storage
Container platforms
Python environments
Monitoring
User management
AI LAB FLOW
UsersResearchers and developers
AI workstationsDevelopment and experimentation
GPU clusterAccelerated compute
FabricHigh-speed data movement
StorageDatasets and shared artifacts
AI CLUSTER FLOW
User / developerDevelopment access and job submission
Login nodeControlled entry point and tools
Scheduler / KubernetesScheduling and orchestration
GPU computeAccelerated model workloads
Fabric + storageData movement and shared datasets
AI CLUSTER
GPU infrastructure for shared AI workloads.
Design around accelerator density, dataset throughput, multi-user scheduling, container runtime, monitoring and operational isolation.
Plan an AI cluster ↗HPC CLUSTER
Scale scientific workloads across CPU and GPU resources.
Typical designs include login and management nodes, compute nodes, Slurm or PBS, MPI/runtime stacks, shared storage, high-speed networking and monitoring.
LoginAccess and development
SchedulerSlurm / PBS
ComputeCPU / GPU nodes
NetworkEthernet / InfiniBand fabric
StorageShared data and checkpoints
HYBRID AI + HPC
Traditional HPC and AI workloads
in one environment.
Separate CPU and GPU partitions while sharing the right operational services, storage and high-speed fabric.
Researchers→Login→Slurm / Kubernetes→CPU partition+GPU partition→Shared storage
DATACENTER / AI FACTORY
Scale architecture from server
to AI factory.
Plan compute, GPU, networking, storage, Linux, Kubernetes, schedulers, monitoring, automation and scientific applications as integrated layers.
Workstation→Server→Cluster→AI Lab→Datacenter→AI Factory
ARCHITECTURE SUPPORT
Need a design sized to your workload?
Share compute, storage and network requirements using the architecture wizard.