High-performance Storage

What is High-Performance Storage?

File Storage—High Performance (HPS) is a fully managed service that delivers cost-efficient, high-performance, and scalable storage for GPU compute workloads. It is designed to support demanding use cases such as AI/ML training, high-performance computing (HPC), and video rendering, where multiple compute instances need simultaneous access to shared data with low latency and high throughput.

Powered by VAST Data and DDN Storage, two of the leading high-performance file system technologies optimized for NVIDIA GPU clusters, HPS provides exceptional performance—offering throughput of up to hundreds of gigabytes per second and tens of thousands of IOPS to meet the needs of the most data-intensive workloads.

How Does It Work?

Mount HPS to GPU Bare Metal and GPU Virtual Machine services via the compute network (subnet) for seamless integration with your GPU workloads.

Why High-performance Storage?

• Highly Performant and Scalable: Achieve hundreds of GB/s throughput and thousands of IOPS for demanding workloads.

• Seamless S3 Integration: Easily access and transfer data from other S3 buckets with the S3 gateway support.

• Simple and Fully Managed: Launch and manage HPS in a couple of hours, with no need for hardware provisioning or ongoing maintenance.