Skip to main content

Armada.ai Bridge Admin Guide

Overview​

Armada.ai GPU Cloud Management Software (also referred to as Bridge) enables GPU-as-a-service providers to build AI clouds.

This document describes the steps to perform administrative functions of the Cloud Management Software.

Document Note​

note

All product names, trademarks, and registered trademarks mentioned in this document are the property of their respective owners. Use of these names, trademarks, and brands does not imply endorsement or affiliation.

Key Topics Covered​

This guide covers the following areas for AI Cloud (GPU-as-a-service, or GPUaaS) Admins:

  • User personas and roles
  • Bridge installation and login
  • Infrastructure discovery / import and management
  • Dashboard overview
  • RBAC (Roles and Permissions)
  • Catalog management
  • Tenant creation and management
  • User management
  • Resource quotas and pricing

Armada Bridge — Admin Overview

Purpose​

Armada Bridge is GPUaaS software that provides secure multi-tenancy, IaaS, PaaS, and AIaaS functionality.

The platform supports three personas:

  • Neocloud/NCP (NVIDIA Cloud Partner) Admin, also referred to as Super Admin
  • Tenant Admin
  • Tenant User

For the Super Admin, Armada Bridge acts as a cloud management and control plane for GPU-based AI infrastructure, enabling providers to operate their GPU fleet as a secure, on-demand, multi-tenant AI cloud.

Bridge provides a single system of record to discover, configure, slice, operate, and monetize GPU infrastructure across:

  • Bare metal
  • Virtualized environments
  • Platform services

All these operations are performed while maintaining hard tenant isolation and high GPU utilization.

What the Admin Owns​

As an Admin, you retain full ownership and control over:

  • Physical infrastructure (GPU servers, networking, storage)
  • Tenant lifecycle and isolation boundaries
  • Service catalogs (BM, VM, PaaS, AI services)
  • Quotas, policies, and monetization controls
  • Observability, security, and compliance

Bridge orchestrates and automates your infrastructure — it does not replace vendor tooling or standards.

Core Admin Capabilities​

1. Infrastructure Discovery/Import & Lifecycle Management​

  • Automated discovery of GPU compute nodes, switches, and fabrics
  • Alternatively, import of infrastructure (without discovery)
  • Topology validation against intended design
  • Day-0 / Day-1 / Day-2 lifecycle automation
  • Fabric underlay configuration
  • Unified control across Ethernet (Spectrum-X) and InfiniBand

2. Unified Multi-Tenancy with Hard Isolation​

Isolation enforced across:

  • GPU and CPU allocation
  • East-West and North-South networking
  • External storage access
  • External gateways and firewalls

Isolation is equivalent to physical separation — not soft namespace-only controls.

Supported models:

  • Dedicated bare-metal tenants
  • Virtualized GPU tenants
  • Fractional GPU (MIG) tenants

3. Service & Offering Definition​

Admins define cloud offerings:

  • Bare Metal-as-a-Service (BMaaS)
  • Virtual Machine-as-a-Service (VMaaS)
  • Kubernetes-based PaaS
  • Job scheduling (native, SLURM, Jupyter on KAI)
  • Model serving and Inference services
  • Fine-tuning services

All services are exposed through a unified API and UI.

4. Tenant Lifecycle & Governance​

  • Create and manage tenant accounts
  • Assign quotas, policies, and RBAC
  • Control service access per tenant
  • Enable identity federation (tenant-owned IdPs)
  • Define billing parameters

5. Observability, Security & Compliance​

  • Single-pane visibility across tenants and infrastructure
  • Per-tenant and per-service usage metrics
  • Activity tracking and auditing
  • Enterprise authentication, authorization, and MFA
  • Production-grade S3P (Stability, Scalability, Security, Performance)

6. Monetization & ROI Optimization​

Increase GPU utilization through:

  • Dynamic allocation and de-allocation
  • GPU virtualization and sharing
  • Managed Kubernetes with autoscaling
  • Intelligent job scheduling
  • Jupyter Notebook support
  • Dedicated and multi-tenant model serving

Avoid vendor lock-in through heterogeneous support:

  • GPU vendors: NVIDIA, AMD
  • Server OEMs: Cisco, HPE, Dell, Supermicro
  • Storage vendors: VAST, DDN, WEKA
  • Networking vendors: NVIDIA Spectrum-X, BF-3, Cisco Nexus 9000, F5 Big IP Next
  • Kubernetes distributions: Red Hat OpenShift, Canonical Kubernetes, SUSE Rancher/RKE

Monetize idle capacity via:

  • Federated marketplaces (e.g., NVCF)
  • Default billing system (replaceable with provider billing)

Admin Value Proposition​

With Armada Bridge, an NCP Admin can:

  • Convert static GPU infrastructure into a revenue-generating AI cloud
  • Support multiple business models on the same fleet
  • Reduce operational overhead through automation
  • Maintain strict tenant isolation without bespoke engineering
  • Maximize ROI on high-cost GPU assets
Last updated on