PanFS
vdura.com
- Linux
- Web
- API
- Self-hosted
What PanFS is
PanFS is the parallel file-system foundation carried into VDURA Data Platform, a software-defined storage platform for AI and high-performance computing. It combines parallel file-system performance with object-storage resilience and cost efficiency. VDURA describes high-throughput storage for AI training, low-latency inference, and persistent key-value cache and session context. The platform is described as scaling linearly to thousands of nodes and tens of thousands of clients, with stated performance of up to 2.7 TB/s per rack. Context-Aware Tiering moves data among NVMe, SSD, and object tiers according to access patterns. REST, Kubernetes CSI, CLI, and S3-compatible interfaces are provided, alongside DirectFlow, NFSv3, SMB/CIFS v3.1, and S3 access. MLEC and client erasure coding support resilience, while VDURA states up to 12 nines of durability. AES-256 protects data at rest and in flight across tenant paths. Multi-tenancy includes per-tenant quality of service, namespaces, and access policies.
Who it suits
It is aimed at Neoclouds, AI factories, enterprises, and HPC environments needing large-scale storage. Hybrid deployment and open hardware support organizations choosing their infrastructure.
Good to know, and what to watch
- Scales to thousands of nodes
- Up to 2.7 TB/s per rack stated
- Context-Aware Tiering across storage tiers
- S3-compatible and Kubernetes interfaces
- AES-256 encryption across tenant paths
- Pricing is available only on request
- Designed for specialized AI and HPC workloads
Plans
Every plan and priceVDURA Data Platform
Pricing on request
The verdict
VDURA Data Platform carries PanFS into a broad AI and HPC storage system with tiering, multi-tenancy, and several access interfaces. Buyers must contact sales for pricing and evaluate the platform against their infrastructure needs.
What the maker publishes
- Architecture
- The platform combines parallel file system performance with object storage resilience and cost efficiency in a unified system. source
- Encryption
- VDURA states that AES-256 encryption protects data at rest and in flight across every tenant data path. source
- Founded
- 2001 source
- Hardware
- VDURA describes its platform as open hardware, letting customers choose preferred hardware vendors. source
- Headquarters
- Pittsburgh, Pennsylvania, United States source
- Headquarters
- VDURA lists its headquarters at 1501 Reedsdale St., Suite 1001, Pittsburgh, Pennsylvania 15233. source
- Interfaces
- The platform provides REST, Kubernetes CSI, and CLI interfaces, plus an S3-compatible object storage interface. source
- Multi-tenancy
- The platform supports per-tenant quality of service, namespaces, and access policies. source
- Performance
- VDURA states performance of up to 2.7 TB/s per rack. source
- Pricing availability
- The opened maker pages direct visitors to contact sales and do not state a product price. source
- Product identity
- VDURA Data Platform carries forward the PanFS parallel file system foundation as software-defined storage for AI and HPC workloads. source
- Resilience
- The platform uses MLEC and client erasure coding, and VDURA claims up to 12 nines of durability. source
- Scalability
- The platform is described as scaling linearly to thousands of nodes and supporting tens of thousands of clients. source
- Support
- VDURA advertises global technical support available 24/7/365. source
- Target users
- The maker identifies Neoclouds, AI factories, enterprises, and HPC environments as intended users. source
- Tiering
- Context-Aware Tiering automatically moves data across NVMe, SSD, and object tiers according to access patterns. source
- Use cases
- The maker describes high-throughput storage for AI training, low-latency access for inference, and persistent KV cache and session context. source
Checked against the maker's pages on 7 Oct 2026.