This section will follow the process of deploying an application to an edge system.
This deployment will use the application DeepStream - Intelligent Video Analytics Demo, which is hosted in the NGC Accelerated Software catalog for this exercise. This Intelligent Video Analytics demo showcases Fleet Command’s management capabilities and the capabilities of GPU-enabled systems for real-time AI-based video and image understanding and multi-sensor processing. This demo combines the Deepstream SDK with a sample model and video source and a media server to provide a web UI for viewing the analyzed video into a single Helm chart easily deployed from Fleet Command. For more information, please refer to the Intelligent Video Analytics Demo.
Before you begin, ensure that you have done the following:
Created a Location in Fleet Command.
Successfully added a System to that Location.
The System status is shown as online and paired with Fleet Command.
To begin, find the application in the NGC Accelerated Software section.
Navigate to the NGC Accelerated Software Section and search for the application.
NoteConfirm you are searching for the application Helm chart and not the container, model, or collections.
Once you have found the application, click on the application name for additional information.
Here you will find the required information to add this application into Fleet Command. Be sure to record the application Name, Version, and Helm Chart URL.
When you click on the Fetch Version, the helm fetch command is copied and looks like below. Extract the Helm fields from the fetch URL as per below.
Helm Chart Name: video-analytics-demo
Helm Chart Version: 0.1.7
Helm Repo URL: https://helm.ngc.nvidia.com/nvidia
Once you have the information listed above, navigate to the Fleet Command Application page to add the application.
Click Add Application.
Confirm display name and description are set.
Once the application is added, it will display on the Fleet Command Application page.
Navigate to the Create Deployment page and fill out the required fields.
Deploy the application.
Additional Configuration for Multi-Instance GPU
Additional configuration may be required for MIG, refer to the Multi-Instance GPU configuration section for more details.
Below is an example Application Configuration for an A100 GPU with two 3c.20gb MIG partitions, from available two MIG partitions on the system below configuration using a MIG partition.
resources:
limits:
nvidia.com/mig-3g.20gb: 1
Once the application has been deployed successfully, you can access the application using the Remote Application Access feature from Fleet Command. For more information on using this feature, refer to Remote Application Access.
Congratulations! You have deployed your first application on Fleet Command.