VAST Data and NVIDIA’s turnkey, petabyte-scale infrastructure solution delivers unrivaled performance for your GPU-intensive and storage-intensive workloads including, Artificial intelligence (AI) and Machine Learning (ML), High Performance Computing (HPC), and Big Data Analytics. These workloads present a new set of storage challenges that you never had to deal with. Legacy NAS storage systems simply cannot deliver adequate bandwidth for powerful NVIDIA GPUs, while parallel file systems are extremely complex.
Enter VAST Data. With VAST Data, you get the performance and scalability of a parallel file system, the simplicity of an all-flash NAS appliance—and the economics of HDD-based archive storage. VAST Data’s Universal Storage consolidates all your data on a single tier of flash, accelerating your most demanding AI applications, shortening deep learning training time, and unlocking previously impossible insights.
Scale to terabytes per second (TB/s) of bandwidth, with millions of IOPS over Ethernet or InfiniBand
VAST Data Universal Storage installs like an appliance with no client software dependency or complexity
All the storage you need for 1/10 the cost of traditional flash storage
VAST server pooling capability provides dedicated QoS for competing applications
NFS over RDMA provides over 10GB/s of mount performance over Ethernet or IB: scale multiple mounts in a single client to drive 100GB/s+ of single client bandwidth
Featured Solution
Jointly designed, built, and tested by NVIDIA and VAST Data, this reference architecture provides enterprises with a turnkey petabyte-scale AI infrastructure solution that is designed to significantly increase storage performance for AI use cases such as large-scale training of conversational AI models and petabyte-scale data analytics.
learn moreCustomers are quickly realizing that the complexities of parallel file systems aren’t worth the hassle when NFS can be equally performant. VAST data has developed several NFS client-side enhancements to deliver best-in-class storage performance for your GPU servers. Among the most exciting is VAST’s collaboration with NVIDIA to power their new GPUDirect™ Storage (GDS) ready systems.
Remote Direct Memory Access (RDMA) enables VAST Data’s Universal Storage to bypass client-side CPU and memory for more agile data transfer efficiency. This results in higher throughput and lower latency, while preserving the operational simplicity of this industry standard storage protocol.
To meet the extreme throughput requirements of NVIDIA’s DGX A100, VAST Data introduced support for NFS multipathing that aggregates client-side network ports access to a single file system. NVIDIA benchmarking showed a VAST Data storage cluster deliver a staggering system throughput of more than 160GiB/s to a single host.
Together with NVIDIA, VAST Data extended NFS support for NVIDIA GDS, creating an expedited path for direct memory access (DMA) transfers between storage and GPU memory—without burdening CPUs. The result is decreased latency, increased storage bandwidth, and dramatically lower CPU utilization.
VAST Data provides the ideal architecture to accelerate, scale and simplify DGX-A100 clusters. With our new support for GPUDirect®, customers can achieve full bandwidth from a single NFS mount point to unleash the potential of AI without HPC storage complexity.
Watch the Nvidia GPUDirect Webinar featuring VAST's resultsVAST Data provides the ideal architecture to accelerate, scale and simplify DGX-2 AI computers. With our new support for GPUDirect, customers can achieve over 85GB/s of bandwidth from a single NFS mount point to unleash the potential of AI without HPC storage complexity.
Watch the Nvidia GPUDirect Webinar featuring VAST's resultsFor any Nvidia AI GPU, VAST Data provides the ideal architecture to accelerate, scale and simplify A100-based computing. With our new support for GPUDirect, customers can achieve up to 25GB/s per GPU from a single NFS mount point to unleash the potential of AI without HPC storage complexity.
Watch the Nvidia GPUDirect Webinar featuring VAST's results