Velocity • Durability
Velocity • Durability
DirectFlow high-performance parallel Client
RDMA CPU-bypass data paths
S3 protocol support
Native Kubernetes CSI plug-in
VeLO: Velocity Layered Operations
Elastic Metadata Engine, up to 20x
VPODs: Virtualized Protected Object Devices
Multi-Level Erasure Coding
Context-Aware Tiering
SMR HDD optimization
Multi-tenancy with per-tenant VLANs and keys
REST API and infrastructure-as-code providers
Prometheus metrics, Grafana dashboards
End-to-end AES-256 encryption, KMIP
Data compression in each Storage Node
Non-disruptive expansion and upgrades
Ethernet and InfiniBand
VDURACare Premier: 10-year drive replacement
Flash Nodes form the performance-driven foundation of the data plane. Each 1U node holds 12 NVMe SSDs from 3.84 TB to 122.88 TB and delivers up to 75 GB/s. Configure as all-NVMe flash for maximum performance, or pair with HDD capacity expansion for economical bulk storage. RDMA moves data between GPU servers and storage with no CPU in the path.
The distributed, flash-optimized metadata engine coordinates file operations across the realm. The V12 Elastic Metadata Engine scales metadata performance across Director Nodes as the system grows.
VPODs™ (Virtualized Protected Object Devices) isolate failures to one virtual device, not the server. Per-file client-side erasure coding (Dual Parity, Single Parity or Striped Mirror) combines with optional Multi-Level Erasure Coding, so durability increases with scale.
Data lands on flash and automatically tiers to HDD based on size, access pattern and temperature, keeping the flash tier lean and hot. Active datasets, checkpoints and model files stay on NVMe; completed work drains to high-density HDD without administrator intervention.
Every node is managed as one realm from a single GUI, CLI or API. Prometheus-compatible metrics feed Grafana dashboards. Data compression runs in each Storage Node with zero client-side impact.
| 42U Rack Configuration Options (not including switches) | |
|---|---|
|
1U Director Node (x3 minimum) |
|
|
VCH-5000-D1N |
Director Node |
|
1U Flash Node (x7) |
|
|
VCH-5000-F1N |
Flash-only, 12 x NVMe SSD (3.84 to 122.88 TB) |
|
VCH-5000-F1N |
Flash with HDD capacity expansion support |
|
4U JBOD, 78 or 108 Drive (x7) |
|
|
VCH-5000-J108-3456 |
108 x 32 TB HDDs |
|
VCH-5000-J108-3240 |
108 x 30 TB HDDs |
|
VCH-5000-J108-2592 |
108 x 24 TB HDDs |
|
VCH-5000-J108-1728 |
108 x 16 TB HDDs |
|
VCH-5000-J78-2496 |
78 x 32 TB HDDs |
|
VCH-5000-J78-2340 |
78 x 30 TB HDDs |
|
VCH-5000-J78-1872 |
78 x 24 TB HDDs |
|
VCH-5000-J78-1248 |
78 x 16 TB HDDs |
The VDURA® Data Platform V5000™ Certified Platform is engineered for AI and HPC pipelines that demand relentless GPU feed rates. Built on industry-standard servers, V5000 pairs NVMe flash with optional HDD capacity expansion in a Mixed Fleet architecture: flash sized for performance, HDD for capacity, and a cost-balanced growth path from a pilot program to production on the exabyte scale.
V5000 runs VDURA Data Platform V12, VDURA’s flash-tuned parallel file system, streaming multiple terabytes per second per rack from a single global namespace. V12 delivers 20% or more additional throughput, 20x metadata acceleration and a 20% or greater reduction in cost per TB at the same performance level versus V11, with a minimum downtime in-place upgrade. Working with the DirectFlow® Client, VDURA offers parallel redundant data paths that scale linearly, safeguard data with enterprise-class durability and keep day-to-day management simple.
Director Nodes anchor the VDURA control plane, running the flash-optimized VeLO™ (Velocity Layered Operations) metadata engine that coordinates billions of file operations and keeps AI pipelines saturated. Each 1U server combines high-core CPUs, plentiful DRAM and dual 200 or 400 Gb/s links, and the V12 Elastic Metadata Engine scales metadata performance dynamically across nodes. Toolless service and redundant power maximize uptime.
Two 4U, 19-inch capacity expansion enclosure options are available, fully populated with 16, 24, 30 or 32 TB HDDs: one with 78 drives for one-meter deep racks and one with 108 drives for 1.4-meter racks. V12 SMR HDD optimization unlocks 25 to 30% more usable capacity per rack. Both include two Titanium-level redundant power supplies and connect over SAS 4 (24G).
| V5000 Performance Specifications | |
|---|---|
|
Throughput |
Up to 75 GB/s per Flash Node Up to 2.7 TB/s per 42U rack, all-flash; 200 GB/s with capacity expansion |
|
IOPS |
Up to 1.2M per Flash Node; 3.6M for a minimum filesystem Up to 45M per 42U rack, all-flash; 9M with capacity expansion |
|
Capacity* |
Up to 17.7 PBe per 42U rack, all-flash; 26 PBe with capacity expansion |
|
Data Durability |
Up to 8 nines, increasing with scale |
|
Data Availability |
Six nines and above in production environments |
|
Protocols |
DirectFlow Client (POSIX), S3, CSI; RDMA |
|
Security |
End-to-end AES-256, KMIP key management, per-tenant keys, RBAC |
* Effective capacity after data compression
Tenants are isolated with per-tenant VLANs, per-tenant AES-256 keys (BYOK), and per-tenant telemetry with hourly metering. The REST API handles tenant lifecycle, quota and billing automation: three calls create a tenant, assign capacity and provision a volume that is mountable in seconds, with consumption data granular enough to invoice from.
Configure from 100% flash to 98% HDD in one namespace. Scale out with additional Flash Nodes for throughput and IOPS, or additional capacity expansion enclosures for bulk storage, without re-architecture. V12 upgrades in place with zero downtime.
| V5000 Technical Specifications | |||||
|---|---|---|---|---|---|
|
|
VCH-5000-D1N |
VCH-5000-F1N |
VCH-5000-J78-* |
VCH-5000-J108-* |
|
|
Description |
Director Node server, 1U |
Flash Node server, 1U |
78-drive JBOD, 4U |
108-drive JBOD, 4U |
|
|
Performance |
N/A |
Up to 1.2M IOPS Up to 75 GB/s |
Up to 13.2 GB/s |
Up to 17.8 GB/s |
|
|
Media |
12 x NVMe SSD slots |
12 x NVMe SSD, 3.84 to 30.72 TB |
78 x HDD, 16 to 32 TB |
108 x HDD, 16 to 32 TB |
|
|
Connectivity |
2 x NDR/NDR200 or 400/200/100 GbE, RoCE v2 |
4 x 200 Gb or 2 x 400 Gb, InfiniBand or Ethernet, RoCE v2 |
SAS 4 (24G) to Storage Node |
SAS 4 (24G) to Storage Node |
|
|
Dimensions W x D x H (without bezel) |
17.5 x 33 x 1.75 in. 45 x 84 x 5 cm |
17.5 x 33 x 1.75 in. 45 x 84 x 5 cm |
17 x 32 x 7 in. 43 x 81 x 18 cm |
17 x 41 x 7 in. 43 x 104 x 18 cm |
|
|
Weight with Drives |
35 lbs. / 15.9 kg |
38.8 lbs. / 17.6 kg |
207.2 lbs. / 94.5 kg |
263.2 lbs. / 119.4 kg |
|
|
AC Power Input |
200 to 240 V, 50/60 Hz, C14 inlet (all models) |
|
|
|
|
|
Power Consumption (typical / max at plug) |
636 W / 981 W |
742 W / 1158 W |
839 W / 1155 W |
1081 W / 1448 W |
|
|
Thermal Output (typical / max) |
2647 / 4097 BTU |
2532 / 3951 BTU |
2862 / 3940 BTU |
3688 / 4940 BTU |
|
|
Power Supplies |
1+1 Titanium |
1+1 Titanium |
1+1 Titanium |
1+1 Titanium |
2 — Self-encrypted drives
3 — Dual-actuator HDDs only