(FORBES)–VDURA, Vdura says that it provides a software-defined data infrastructure platform built for AI and high-performance computing, HPC. Their software combines the scalable, linear performance of a parallel file system with the resilience and cost efficiency of object storage, providing a unified global name space with one control plane and one data plane.
Panasas rebranded itself as Vdura on May 7, 2024 as it made a strategic switch from selling proprietary hardware to focusing on software-defined storage with a subscription-based business model for AI and high-performance computing.
Panasas was founded in 2000 by Garth Gibson, co-inventor of RAID and the creator of the first Linux-based parallel file system, PanFS. In 2004 Panasas started to produce high-performance storage systems. PanFS was the basis of the company’s switch to focusing on software subscriptions. Garth Gibson returned to Vdura as Chief Technology and AI officer in September 2025.
To support AI workloads sustained throughput at scale is essential to avoid starving GPUs and wasting that investment. AI also requires expanding data metadata capacity and performance to quickly access the data. Vdura says that they accelerate in all parts of the AI pipeline rather than in certain parts like their major competitors. The figure below shows these various parts including data ingesting, model loading, training, check pointing, fine-tuning and inference.
The company says that their Vdura Data Platform, VDP, can turn 6-thousands of storage servers into a high-performance, resilient and durable data platform. It is resistant to many failures of both devices and nodes, is easy to deploy and manage and provides seamless integration of NVMe for a fast access layer and optional hard disk drives, HDDs, for cost efficient storage capacity.
In particular, the company can use SATA rather than serial attached SCSI, SAS, HDDs, which are widely used by hyperscalers for object storage as part of their software defined package. The figure below shows recent costs of various SSDs using flash memory versus SAS and SATA HDDs.
Vdura says that NVMe flash + HDD architectures reduces the long-term costs versus all-flash competitors, especially using commodity SSDs and HDDs. The Vdura-bility guarantee and Vdura Care Premier provide guaranteed SLAs for availability and durability plus 10-year cover with device replacement.
By enabling fast and reliable performance using a combination of SSDs and HDDs Vdura can offer a flexible and adaptable ratio of HDDs and flash in a system as shown below. The company told me that an 80%/20% HDD to SSD mix is best for most customers.
Adaptive performance scaling with Vdura (VDURA)
To get the most use of GPUs the storage system that supports it must have high availability, often referred to as the percent uptime as the number of nines before and after the period, e.g. 99.999% uptime is 5-nines.
Also, the data used for AI must be able to last with no loss or corruption over time. Although storage is about 20% of the spending for AI it enables the use of the 50% server/GPU investment. Availability and durability reduce the incidence and impact of downtime.
In September Vdura announced a validated reference architecture with AMD designed to eliminate data bottlenecks and simplify large-scale deployments for AI and HCP. This combines Vdura V5000 storage with AMD Instinct MI300 series accelerators.
Vdura says it offers lower capital and operating costs and with high storage availability and durability for demanding at-scale AI applications using combinations of SSDs and HDDs.