Never let your GPUs wait.
The data platform for AI factories.
VDURA Data Platform is a software-defined data platform built for AI and HPC. It automatically places data across NVMe flash and HDD within a single namespace, keeping GPUs continuously fed while significantly reducing storage costs. Built on PanFS, a parallel file system with more than 20 years of proven history originally developed by Panasas (now VDURA), it runs in more than 1,000 production environments worldwide.
What is VDURA Data Platform?
Fast
No more waiting GPUs
Data is delivered in parallel from multiple nodes, achieving up to 2.7 TB/s of bandwidth per rack.
Cost-Efficient
Lower cost with standard hardware
No proprietary appliances: the system is built on commodity servers and standard HDDs/SSDs. Scale performance with flash and capacity with HDD, adding only what you need.
Simple
Manage everything from one place
No matter how many nodes you add, the entire system is managed from a single GUI or CLI.
VDURA Data Platform is a storage platform that combines VDURA’s parallel file system software with industry-standard hardware from multiple vendors. Because it runs on commodity servers, you are never locked into a single vendor’s proprietary hardware.
The software handles all data placement, protection, and acceleration, so there is no need for expensive specialized hardware. The system uses proven, certified products: servers from vendors such as Supermicro and Dell, Seagate HDDs, and SSDs from vendors such as KIOXIA and Solidigm.
This combination of software and standard hardware delivers both the speed of flash and the high capacity and low cost of HDD in a single system. It brings the “flash for performance, HDD for capacity” approach used by the world’s largest hyperscalers to enterprises and research institutions.
Why VDURA and why now?
Three challenges with conventional storage architectures
As investment in GPUs grows, more and more organizations are finding that storage is holding their GPUs back. In AI infrastructure especially, the bottleneck is no longer compute. It is data.
Separate systems for performance and capacity
Pairing fast flash with low-cost HDD or object storage from different products means two management consoles, two support contacts, and data that must be copied between them.
GPUs stall when hardware fails
When a drive or node fails, the system is consumed by rebuilding data, and the flow of data to the GPUs slows to a crawl.
The high cost of proprietary hardware
Proprietary storage appliances depend on specialized components for redundancy, driving up costs every time you expand.
VDURA solves all three on a single platform
VDURA addresses all three challenges with one control plane, one data plane and one namespace.
Unified in a single system
NVMe flash and HDD are managed under a single namespace. There is no need to copy data between tiers.
Continuous data delivery, even during failures
Client-side erasure coding keeps data flowing to your GPUs, even when a failure occurs.
Reliability without proprietary hardware
Data protection is handled in software, so commodity servers and drives are all it takes.
The three components of the V5000
| V5000 Director Node | The Director Node is the command center of the system, responsible for all metadata processing. It runs VeLO, a flash-optimized metadata engine that manages billions of file operations so your AI pipelines never slow down. Each compact 1U server is equipped with high-core-count CPUs, large DRAM capacity, and dual 200 Gb/s network connections for extremely low-latency coordination. Data is protected by instant secure erase and optional end-to-end encryption. Tool-free serviceability and redundant power supplies help maximize uptime in demanding AI and HPC environments. |
| V5000 Flash Node | Flash Nodes power the data plane and are used solely to store user data. Two configurations are available. For maximum performance, choose an all-NVMe flash configuration. For large-scale storage that balances performance and cost, combine NVMe flash with HDD capacity expansion. Either way, you get infrastructure tuned to your workload. |
| V5000 Capacity Expansion Enclosure | The V5000 offers two capacity expansion enclosure models. Choose a high-density model with 78 HDDs or an ultra-dense model with 108 HDDs. Both are 4U, 19-inch rack-mount units that ship with every drive bay filled. The 4U78 fits standard 1-meter-deep racks, while the 4U108 is designed for 1.2-meter-deep racks. A wide range of HDD capacities is supported, as shown below. Each enclosure also includes two redundant 80 PLUS Titanium power supplies. |
Tune the balance of speed and capacity
With the V5000, you can choose any mix from all-flash to as much as 98% HDD. Need more speed? Add Flash Nodes. Need more capacity? Add expansion enclosures. And you can scale without taking the system offline.
Key use cases
AI training
Load massive training datasets at high speed and write checkpoints in a fraction of the time.
AI inference and RAG
Serve models and embeddings with low latency. Use them directly from the same data space used for training.
Long-term memory for AI agents
Store KV cache and session data outside GPU memory to support long conversations and extended workloads.
HPC and R&D
A computing foundation for research institutions and manufacturers running data-intensive simulation and analysis.
Deployment and support
VDURA Data Platform is deployed as a software subscription combined with certified hardware. Long-term support plans are also available for the hardware. As an authorized distributor across the Asia-Pacific region, including Japan and India, Pacific Teck supports you at every step, from system design and deployment to ongoing maintenance.
1,000+ deployments worldwide
Chosen by defense organizations, medical research centers, universities, and government agencies where performance and reliability are critical.
Can I use my existing servers and HDDs?
VDURA Data Platform runs on VDURA-certified hardware. Please contact us for details.
Can I start with a small configuration?
Yes. Systems start at six nodes and can be expanded later without downtime.
Can it be used in the cloud?
Yes. It supports on-premises, public cloud, and hybrid environments.
Request a quote or contact us about VDURA Data Platform
Pacific Teck provides pre-sales and post-sales support for VDURA Data Platform across the Asia-Pacific region.
For inquiries, please contact us by email at info@pacificteck.com
PARTNERS
