What is AWS ParallelCluster?
AWS ParallelCluster is an open-source tool for creating and managing High Performance Computing (HPC) clusters in AWS. It automates deployment of complete HPC environments: compute nodes, shared filesystems, job schedulers, and networking.
ParallelCluster enables researchers and engineers to access virtually unlimited computing power without operating their own HPC infrastructure.
Core Features
- Automated Cluster Deployment: Complete HPC environment in minutes
- Auto-Scaling: Compute nodes scale based on job queue
- Slurm Integration: Proven job scheduler for HPC workloads
- Shared Storage: Integration with FSx for Lustre, EFS, and S3
- Spot Instance Support: Up to 90% cost savings with Spot Instances
Typical Use Cases
Scientific Simulations: CFD simulations, molecular dynamics, weather models, and other compute-intensive scientific applications benefit from elastic HPC capacity.
Machine Learning Training: Distributed training of ML models on GPU clusters. Scale from one GPU to hundreds for large models.
Life Sciences: Genomics pipelines, protein folding, and drug discovery with access to specialized instances like p4d for NVIDIA A100 GPUs.
Benefits
- No hardware investment for HPC projects
- Elastic capacity: scale to thousands of cores in minutes
- Pay-per-use with Spot Instances for massive cost savings
- Familiar environment for HPC users with Slurm
Integration with innFactory
As an AWS Reseller, innFactory supports you with AWS ParallelCluster: We help with HPC cluster architecture, migration of on-premises HPC workloads, and optimization for cost and performance.
Typical Use Cases
Frequently Asked Questions
What is AWS ParallelCluster?
AWS ParallelCluster is an open-source tool for creating and managing HPC clusters in AWS. It automates deployment of compute nodes, shared storage, and job schedulers for scientific computing.
Which job schedulers are supported?
ParallelCluster supports Slurm and AWS Batch as job schedulers. Slurm is the standard for traditional HPC workloads, AWS Batch for container-based computing.
What does AWS ParallelCluster cost?
ParallelCluster itself is free. You only pay for AWS resources: EC2 instances, EBS volumes, S3 storage, and network. Spot Instances can save up to 90% on costs.
Which instance types are suitable for HPC?
Hpc6a and Hpc7g offer optimized networking for tightly coupled workloads. P4d and P5 for GPU computing. C6i and C7g for compute-intensive workloads.