Watch the film
LightSpeed Philosophy
not just a product. A machine learning storage philosophy.
As organizations race to develop new AI-based product and service offerings, this new class of computing interacts with data in ways never-before encountered. This raises critical infrastructure planning issues that will challenge how storage has been built for decades.
VAST combines a new hardware offering, blueprints for scalability & forthcoming support for accelerated GPU computing to establish Universal Storage as the platform of choice for intelligent applications.
the light-touch simplicity of nas
the speed to support any ai ambition
Coupled With The Scale & Economics Of VAST’s Award-Winning DASE™ Architecture
Watch the film
New Storage HW for AI Workloads
the storage platform for the next decade of machine intelligence.
-
ideal performance for ai
Designed for AI Workloads with the power of 3,000 HDDs in 2U
-
Unrivaled Flash Capacity
Over 1PB of capacity for training and inference data in every 2U device
-
Scale to Exabytes
Balance your GPU cluster scale with capacity & performance-rich storage
Accelerated NFS for GPUs
unrivaled simplicity and speed
Most enterprise storage systems adhere to the standard NFS/TCP model that has been the core offering of scale-out NAS providers for decades. New AI machines need more than the 2GB/s of read throughput than a standard TCP NFS client can provide. This single-stream, non-RDMA limitation is why organizations abandon NFS in favor of parallel file systems and chose performance over simplicity.
In 2020, VAST has combined support for a number of NFS accelerations that are in the Linux kernel to make it possible to get best-in-class throughput for AI applications and GPU computers without requiring customers to choose complex parallel file systems.
NVIDIA® GPUDirect® enables VAST servers to directly place data in GPU memory, bypassing CPU memory bottlenecks to increase throughput, eliminate CPU overhead and lower latency for GPU I/O.
Up to
40GB/sPer LightSpeed Enclosure
More Than
88GB/svia NVIDIA DGX-2™’s
Scalable Reference Architectures
Born in thE Era of scalable ai clustering.
Universal Storage clusters can scale to meet the need of any AI ambition. These example cluster configurations can help to build balanced architectures that combine the speed needed to keep AI processors working with all-flash capacity needed for deep learning.
Example GPU Cluster Architectures
-
40 gb/s
1 pb*
400k iops
1x LightSpeed Enclosure
1x VAST Server Chassis
Right-sized to provide balanced I/O for up to 16 GPU clients
-
200 gb/s
5 pb*
2m iops
5x LightSpeed Enclosure
5x VAST Server Chassis
Right-sized to provide balanced I/O for up to 80 GPU clients
-
400 gb/s
10 pb*
4m iops
10x LightSpeed Enclosure
10x VAST Server Chassis
Right-sized to provide balanced I/O for up to 160 GPU clients
+
“VAST delivered an all-flash solution at a cost that not only allowed us to upgrade to all-flash and eliminate our storage tiers, but also saved us enough to pay for more GPUs to accelerate our research. This combination has enabled us to explore new deep-learning techniques that have unlocked invaluable insights in image reconstruction, image analysis, and image parcellation both today and for years to come.”
Bruce Rosen, Executive Director
Martinos Center for Biomedical Imaging
- All-Flash for less than the cost of tiered flash & HDD
- Storage savings that provided budget to buy more GPUs
- A simple all-flash NAS platform to eliminate AI application complexity
