Velocity • Durability

VDURA — API
AI SOLUTIONS / API

Storage that sits under your control plane, not beside it.

A REST API control plane handles tenant lifecycle, quota, and billing automation, built for infrastructure-as-code workflows. Your portal and your billing system run the storage lifecycle without a human in the loop.

✓ Infrastructure as code ✓ Native Kubernetes CSI ✓ Prometheus exporter
ONBOARD A TENANT · THREE CALLS, NO TICKETS
POST /api/v1/tenants/
{ "name": "Northwind AI", "short_name": "northwind",
  "network_assignments": [{ "network_id": "vlan-101" }] }
→ 201 · tenant isolated on its own VLAN

PUT /api/v1/tenants/{id}/storagesets/{flash-pool}
{ "quota_bytes": "240TiB" }
→ 200 · tenant capacity assigned

POST /api/v1/volumes/
{ "pathname": "/northwind/train",
  "storageset_id": "flash-pool-a",
  "hard_quota_bytes": "200TiB" }
→ 201 Created · mountable in seconds
Illustrative against the v1 API surface. Every call Bearer-authenticated, RBAC-scoped, and audit-logged. No LUN carving, no rebalance, no change window.
WHERE THE API SITS

Your portal calls it. Your billing reads it. Your pipeline declares it.

A GPU cloud already has a control plane. Storage should be a resource it addresses, not a second console someone has to remember to log into.

YOUR SYSTEMS Customer portal Billing pipeline IaC tooling Kubernetes · CSI VDURA REST API RBAC-scoped tokens immutable audit log CLI and Ops Portal are being rebuilt on this same surface Director control plane · stateless quorum Provisioning volumes · quotas Telemetry per-tenant · health Metering per volume · per storage set One namespace · flash and HDD · every protocol
WHAT THE API COVERS

Five surfaces, and the one that pays your invoices is included.

Provisioning /volumes
· Tenants, volumes, quotas, encryption, and VLAN assignments, each an auditable call
· Snapshots for dataset versioning
· No LUN carving, no rebalance, no change window
Telemetry /metrics
· Per-tenant metrics, health, and events
· Native Prometheus-compatible query API, with pre-built Grafana dashboards
· Per-tenant p99 latency, queue depth, and cache hit ratios
Placement
· Tiering policies and placement rules
· External tiering engines plug into VPOD hooks, so your team can tune placement with us
Metering /volumes/{id}/space
· Consumption per volume and per storage set
· A direct feed into your billing pipeline
· What you invoice matches what the tenant consumed
Access control
· RBAC-scoped sessions: platform-admin, tenant-admin, and tenant-user roles
· Every call lands in an immutable, UTC-stamped audit log
Kubernetes CSI
· Zero-script persistent-volume provisioning through standard Kubernetes APIs
· Dynamic provisioning, online expansion, and snapshots
· Dataset versioning for ML workflows without custom integration work
YOUR INTERFACES

Five ways in, one system underneath.

Every interface resolves to the same control plane. The Ops Portal and CLI are built on the same API surface your automation calls, so nothing is reachable one way and not another.

GUI
VDURA Ops Portal
Health, capacity, tenants, and tickets. A shared view with the VDURA NOC.
SHELL
Management CLI
Full cluster admin, scriptable, and used for everything the portal does.
AUTOMATION
REST API
Provisioning, telemetry, and tenant lifecycle. Built for control-plane and billing integration.
TELEMETRY
Prometheus and Grafana
Prometheus-compatible metrics and logs, with pre-built dashboards.
ORCHESTRATION
Kubernetes CSI
Dynamic volume provisioning, online expansion, and snapshots.
WORKED EXAMPLE · TENANT ONBOARDING, END TO END

Contract signed to first byte written, with nobody filing a ticket.

1
Create
Three calls create the tenant, assign capacity, and provision the volume.
2
Mount
DirectFlow on compute, a CSI PVC binding in Kubernetes, or S3 credentials issued.
3
Observe
The volume auto-appears in Prometheus and Grafana with per-tenant latency, IOPS, and capacity.
4
Bill
A consumption feed per volume flows straight into your billing system.
5
Adjust
Grow a quota or snapshot a volume. All online, with no remount.
Design goal: your portal and billing never need a human in the loop for storage. The tenant lifecycle is fully API-driven.
INFRASTRUCTURE AS CODE

Declare the tenant. Let the platform reconcile it.

The API is built for declarative automation, so storage lands in the same repo and the same review as the rest of your fleet. The native CSI plug-in handles zero-script persistent-volume provisioning, expansion, and snapshots.

Storage reviewed in the same pull request as the compute it serves
Marketplace control plane integration, so a listing can provision real capacity
Configuration management deploys the DirectFlow client fleet-wide from our runbook
Drift is visible, because the desired state lives in your repo rather than in someone’s memory of a console
MAIN.TF
resource "vdura_tenant" "northwind" {
  name       = "Northwind AI"
  short_name = "northwind"

  network_assignment {
    network_id = "vlan-101"
  }
}

resource "vdura_tenant_storageset" "nw_flash" {
  tenant_id     = vdura_tenant.northwind.id
  storageset_id = "flash-pool-a"
  quota_bytes   = "240TiB"
}

resource "vdura_volume" "nw_train" {
  pathname          = "/northwind/train"
  storageset_id     = "flash-pool-a"
  hard_quota_bytes  = "200TiB"
}
Illustrative. Talk to us for the current provider schema.
TELEMETRY AND METERING

The same feed answers "is it healthy" and "what do we invoice".

Native
PROMETHEUS QUERY API
Native exporter, no sidecar to build, with pre-built Grafana dashboards.
Per volume
CONSUMPTION METERING
Consumption data granular enough to invoice from, per tenant.
Per tenant
NOISY-NEIGHBOR SIGNALS
p99 latency, queue depth, and cache hit ratios attributable to a specific tenant.
Immutable
AUDIT LOG
Every API call recorded with a UTC timestamp, retained for compliance review.
WHAT THIS MEANS FOR THE BUSINESS

Automation is what lets tenant count grow faster than headcount.

Faster Time To Revenue
Storage stops being the ticket that holds up a GPU contract. Tenants onboard in minutes through the API rather than in change windows.
A Lean Team Runs An Exabyte
Tenant count can grow faster than headcount only if every step of the lifecycle is a call rather than a person.
Billing That Matches Usage
Consumption metering per volume feeds your invoicing directly, so revenue tracks consumption instead of an estimate.
No Second Console
Your operators live in your control plane. Storage that answers to it does not add a tool for someone to forget.
Security Review Passes
RBAC-scoped access and an immutable audit log answer the questions that otherwise stall an enterprise deal.
Built On Your Standards
Standard infrastructure-as-code workflows, Kubernetes CSI, and Prometheus. Nothing here asks your team to learn a proprietary stack.

Learn more about automating storage with the VDURA API.