Breadcrumbs

NVIDIA nVector Integration

Overview

Login VSI, in collaboration with NVIDIA, has developed an integration between Login Enterprise and NVIDIA nVector that leverages GPU and virtual GPU (vGPU) performance testing and analysis technology. This integration enables you to collect round-trip graphics latency data and image quality scores from your VDI environments, upload them to Login Enterprise as platform metrics, and use them to optimize GPU-based configurations for performance and cost-efficiency.

Latency measurement is a production feature of Login Enterprise. SSIM-based image quality scoring is available as a usable preview - functional and customer-usable, but not yet a built-in Login Enterprise feature. For full setup and usage details, refer to the installation guide PDF in the GitHub repo.

Requirements

  • Login Enterprise Appliance v5.13.6 or higher (which has Platform Metrics enablement).

  • Supported VDI stacks: Citrix, VMware/Omnissa (Horizon), RDS, Azure Virtual Desktop.

  • NVIDIA GPU (or vGPU) present on the Launcher machine - nVector workloads require GPU-enabled Launchers.

  • nvector-agent.exe on the Appliance and each Launcher. v2.0.0 is required for SSIM preview functionality. Obtain from Blair Parkhill (b.parkhill@loginvsi.com).

  • ssim-tool.exe on the Launcher (v1.1.0, required for SSIM preview mode only). Available from Blair Parkhill (b.parkhill@loginvsi.com).

  • Launcher access to the Login Enterprise API (for metric upload).

  • Following the Quick Start section below.

Quick Start

  1. git clone https://github.com/LoginVSI/nVector

  2. Upload the agent to the Appliance

    1. Via a tool like WinSCP, place nvector-agent.exe v2.0.0 in the Appliance script content folder at loginvsi/content/scriptcontent/nvector-agent.exe.

  3. Create an Environment in Login Enterprise and note the EnvironmentId (UUID). To learn more, see Creating an Environment.

  4. Upload the workload

    1. Add NVIDIA_nVector_Desktop_Prepare.cs from the repository as a custom application/workload in Login Enterprise. For details, refer to Importing Application Scripts.

  5. Prepare Launchers (per Launcher you will use for latency measurements):

    1. Copy the following to the Launcher (recommended path is C:\ProgramData\NVIDIA Corporation\nVector): nVector_Client_Prepare.ps1, nVector_SSIM_Daemon.ps1, nVector_Startup_Script.cmd, nvector-agent.exe (v2.0.0), ssim-tool.exe (required for SSIM mode).

      1. Note: nVector_Client_Prepare.ps1 and nVector_SSIM_Daemon.ps1 must be in the same folder. Do not invoke nVector_SSIM_Daemon.ps1 directly - it is spawned automatically by nVector_Client_Prepare.ps1.

    2. Edit nVector_Client_Prepare.ps1 and set these values in the User Config section, or pass them on the command line:

      1. BaseUrl > your Login Enterprise appliance base URL.

      2. ConfigurationAccessToken > API token with Configuration access level. For more information, consult Adding a System Access Token.

      3. EnvironmentId > the UUID from step 3 (for latency metrics).

      4. TargetHost > hostname of the target/desktop machine.

      5. SsimEnvironmentId > UUID of a second LE environment for SSIM metrics (recommended - see installation guide for details on single vs. two environment setup).

      6. Mode > both (default), latency, or ssim.

        1. For the full parameter reference, see the installation guide PDF in the repository.

    3. Place nVector_Startup_Script.cmd in the user Startup folder (like "shell:common startup" or create a scheduled task to run at logon, such as:

      1. powershell.exe -ExecutionPolicy Bypass -File "C:\temp\nvidia\nVector_Client_Prepare.ps1"

  6. Create a Test in Login Enterprise (Load Test or Continuous):

    1. Use the Environment created earlier.

    2. Add the nVector_Desktop_Prepare workload to a session.

    3. Important: set the connector/session to run full-screen on the Target VM (nVector captures the top-most full-screen session - see note below).

  7. Run a short smoke test (60–120s) to confirm datapoints are uploaded to Platform Metrics and the CSV uploader writes files.

Now you're ready to scale up and run further actual testing.

What You’ll Get

  • Scatter/time charts of endpoint graphical latency (ms) in Login Enterprise Platform Metrics.

  • SSIM-based image quality scores in Login Enterprise Platform Metrics (preview feature - requires nvector-agent.exe v2.0.0 and ssim-tool.exe v1.1.0).

  • Per-launcher CSV/JSON exports of latency and SSIM data via the included get_nVectorMetrics.ps1.

  • Ability to compare GPU vs no-GPU scenarios and correlate with other Login Enterprise measurements (EUX, VSImax, session metrics, etc.).

What to Check After the First Run

  • Platform Metrics in the Appliance's UI show latency datapoints (scatter/time).

  • If running in SSIM mode: Platform Metrics in the SSIM environment show SSIM score data points.

  • The Launcher folder contains nvector logs and the SSIM daemon log.

  • get_nVectorMetrics.ps1 can pull data for your EnvironmentId and time window. To retrieve latency and SSIM data together, use -EnvironmentIds "latency-env-uuid", "ssim-env-uuid".

Enabling Browser Viewing of Latency Metrics

  1. Open the Login Enterprise web UI and sign in with an account that can view the Appliance UI.

  2. Open the browser Developer Tools:

    • Windows: press F12 or Ctrl + Shift + I

    • macOS: Cmd + Option + I

  3. Select the Console tab in Developer Tools.

  4. Paste the following command at the console prompt and press Enter:

    1. leSetFeatureFlag('platformMetrics', 1)

    2. This will enable the Platform Metrics view, so nVector metrics (latency, SSIM, etc.) become visible in the UI. This feature flag is currently required on all Login Enterprise versions.

  5. Refresh the page (if the metrics view does not appear immediately) and navigate to the Platform Metrics / Results area to view the nVector charts.

How to View and Export Results

  • In the UI: 

    • For Continuous Test: select Platform Metrics > the environment/time window > view Endpoint Latency scatter and timeseries.

    • For Load Test: the measurements will automatically display in the test results

  • Export: Run get_nVectorMetrics.ps1 with -LEApiToken, -EnvironmentId (or -EnvironmentIds for multiple), and either -LastHours or -StartTime/-EndTime to export CSV/JSON. Use -EnvironmentIds "latency-env-uuid", "ssim-env-uuid" to pull latency and SSIM data together in one run. (Optional: -ImportServerCert and -KeepCert for appliances using self-signed/private CA certs.)

    • If you don’t see metrics: Confirm the API token, EnvironmentId, and that the Launcher uploaded CSVs to the Appliance. If you hit SSL/certificate errors, try -ImportServerCert (and ensure -BaseUrl matches the server’s cert hostname).

Important Operational Notes

  • Launcher must be GPU-enabled. nVector requires the Launcher to have an NVIDIA GPU/vGPU — CPU-only Launchers will not generate latency telemetry.

  • Latency is collected only from the top-most full-screen session on a Launcher. If multiple desktop sessions run on a single Launcher, only the active full-screen session will emit latency measurements. Best practice: if possible, run one Target session per Launcher when you require latency metrics.

Troubleshooting: Quick Checks

  • No latency datapoints:

    • Confirm nvector-agent.exe exists and nVector_Client_Prepare.ps1 ran successfully at logon.

    • Verify the test session is full-screen on the Target VM (only full-screen sessions produce latency).

    • Confirm Launcher is GPU-enabled and drivers are up to date.

  • CSV upload failed/missing exports:

    • Inspect local Launcher logs (path set in nVector_Client_Prepare.ps1).

    • Run get_nVectorMetrics.ps1 manually with the EnvironmentId/API token and check for HTTP errors.

  • Platform Metrics UI not visible:

    • Enable the platform metrics UI flag as documented in Enabling Browser Viewing. The flag is currently required on all Login Enterprise versions. Confirm the Appliance is v5.13.6 or newer.

(See the PDF in the repo for log examples, feature-flag commands, and advanced diagnostics.)

Contacts and Support

Appendix: File Reference

  • nVector_Client_Prepare.ps1 - main launcher-side orchestration script. Manages the nVector agent lifecycle, collects latency data, uploads platform metrics to Login Enterprise, and coordinates SSIM processing via the daemon.

  • nVector_SSIM_Daemon.ps1 - SSIM processing daemon. Spawned automatically by nVector_Client_Prepare.ps1 when running in SSIM or both modes. Do not invoke directly. Must be in the same folder as nVector_Client_Prepare.ps1.

  • nVector_Startup_Script.cmd - startup wrapper that invokes nVector_Client_Prepare.ps1 automatically at logon on the Launcher.

  • NVIDIA_nVector_Desktop_Prepare.cs - Login Enterprise workload for the target/desktop machine. Downloads and starts nvector-agent.exe in the desktop role during a test session.

  • get_nVectorMetrics.ps1 - export helper that queries the Login Enterprise API and outputs timestamped CSV and JSON. Supports single or multiple environment IDs in one run.

  • Create-NvidiaGpuSessionMetricGroup.ps1 - one-time setup script that creates a pre-configured NVIDIA GPU session metrics group in Login Enterprise via the API.

  • Add-SessionMetrics.ps1 - interactive script for discovering and registering additional PerfMon and WMI session metrics in Login Enterprise.

  • nVector_Agent_Installation_x.pdf - full installation guide, parameter reference, and troubleshooting guide (in repo).

Additional Resources