33.2e vGPU licensing server: NVIDIA License System (NLS) and license management

📦 Virtualization and Cloud 📖 GPU Virtualization

🧭 Context Introduction

When you use NVIDIA vGPU to share a physical GPU across multiple virtual machines (VMs), each VM needs a valid software license to activate the vGPU feature. Without a license, the vGPU will not function or will run in a degraded mode. The NVIDIA License System (NLS) is the modern, cloud-connected licensing solution that manages these vGPU licenses. As a new engineer, understanding how NLS works and how to manage licenses is essential for keeping your virtualized AI infrastructure operational and compliant.

⚙️ What is the NVIDIA License System (NLS)?

The NVIDIA License System is a software-based licensing platform that replaces older methods like hardware dongles or static license files. It allows you to:

  • Centrally manage vGPU licenses across multiple hosts.
  • Track license usage in real time.
  • Support both online and offline (air-gapped) environments.
  • Automatically renew or release licenses when VMs start or stop.

NLS consists of two main components:

  • NLS Client – Runs on each hypervisor host (e.g., VMware ESXi, KVM) and communicates with the license server.
  • NLS License Server – A dedicated server (physical or virtual) that hosts the license pool and responds to client requests.

🛠️ How License Management Works

🔑 License Types

License Type Description Use Case
Evaluation License Temporary, time-limited license (usually 90 days) Testing vGPU before purchase
Subscription License Time-based license (e.g., 1 year, 3 years) Production environments with ongoing support
Perpetual License One-time purchase, no expiration Long-term deployments (less common now)

🔄 License Flow

  1. A VM with vGPU starts.
  2. The hypervisor’s NLS Client contacts the NLS License Server.
  3. The server checks if a license is available in the pool.
  4. If available, the server leases a license to the VM for a period (default: 24 hours).
  5. The VM continues to renew the lease while running.
  6. When the VM stops, the license is released back to the pool.

📊 NLS Deployment Options

☁️ Online Mode (Cloud-Connected)

  • The NLS License Server connects to NVIDIA’s cloud portal to validate and download licenses.
  • Requires internet access.
  • Simplest to set up — licenses are automatically synchronized.

🏠 Offline (Air-Gapped) Mode

  • The NLS License Server has no internet connection.
  • You manually download a license file (.bin) from the NVIDIA Licensing Portal and upload it to the server.
  • Used in secure or isolated environments (e.g., government, defense).

📊 Visual Representation: NVIDIA Cloud License System (CLS) token checkout

This flowchart demonstrates how client VMs checkout license tokens from license servers to enable CUDA compute.

flowchart LR VM["Guest VM Boot"] --> Request["1. Request token from License Server (DLS/CLS)"] Request --> Validate["2. Validate license entitlement"] Validate --> Expose["3. Expose full GPU capability to CUDA stack"] classDef cpu fill:#eafaf1,stroke:#76b900,stroke-width:2px,rx:6px,ry:6px; classDef memory fill:#f0f7ff,stroke:#3498db,stroke-width:1.5px,rx:4px,ry:4px; classDef system fill:#f1f5f9,stroke:#64748b,stroke-width:1.5px; class Validate cpu; class Expose memory; class VM,Request system;

🕵️ Key Management Tasks

✅ Installing the NLS License Server

  • The NLS License Server is a lightweight application that runs on Linux (Ubuntu, RHEL) or Windows Server.
  • After installation, you configure it with a server name and port (default: 8080 for HTTP, 443 for HTTPS).
  • You then activate the server using a product activation key from NVIDIA.

✅ Configuring the NLS Client on Hypervisors

  • On each hypervisor host, you install the NLS Client package.
  • You point the client to the NLS License Server’s IP address or hostname.
  • The client will automatically request licenses when vGPU VMs are created.

✅ Monitoring License Usage

  • Access the NLS License Server’s web interface (e.g., https://nls-server:8080).
  • You can see:
  • Total licenses available.
  • Licenses currently in use.
  • Which VMs or hosts are consuming licenses.
  • License expiration dates.

✅ Troubleshooting Common Issues

  • "No licenses available" – Check if the license pool is exhausted or if licenses have expired.
  • "License server unreachable" – Verify network connectivity between the hypervisor and the NLS server (firewall rules, DNS).
  • "License lease expired" – Ensure the NLS Client service is running and can reach the server.

📋 Comparison: NLS vs. Legacy Licensing

Feature NVIDIA License System (NLS) Legacy (Dongle/File-based)
Connection Cloud-connected or offline Static file or USB dongle
Scalability Centralized, supports many hosts Per-host or per-dongle limits
License Reclamation Automatic when VM stops Manual or requires reboot
Monitoring Real-time web dashboard No built-in monitoring
Updates Automatic via cloud Manual file replacement

🧩 Best Practices for New Engineers

  • Start with evaluation licenses to test your vGPU setup before committing to production licenses.
  • Use a dedicated VM or small physical server for the NLS License Server — do not run it on a hypervisor host.
  • Back up your license server configuration and license files regularly.
  • Monitor license usage weekly to avoid running out of licenses during critical workloads.
  • Plan for redundancy — in large deployments, run multiple NLS License Servers in a cluster for high availability.

✅ Summary

The NVIDIA License System (NLS) is the modern way to manage vGPU licenses in virtualized AI environments. It replaces old dongles and static files with a flexible, centralized server that can work online or offline. As a new engineer, your main tasks will be installing the NLS server, configuring clients on hypervisors, and monitoring license usage to ensure your vGPU-enabled VMs always have the licenses they need to run AI workloads.