Device Groups
Before deploying an instance of Modzy Edge to a remote device, you'll first need to create a Device Group.
A Device Group is a collection of configuration, Modzy models, registration tokens, and registered devices.
All devices registered under a particular Device Group will inherit that Device Group's configuration, and be predisposed to download and run the models associated with the Device Group upon startup.
Configuration
There are a number of configuration options available to you when creating your Device Group. Some values can be omitted completely, while others are only necessary in certain use cases. The table below describes the available configuration options.
| Configuration Field | Description | Default | Required |
|---|---|---|---|
| data-dir | Modzy Edge will need to store some information on the local disk. This specifies that path in which to do so | ~/.modzy or /opt/modzy | False |
| runtime | The container runtime utilized for running the model containers | Docker | True |
| model.registry.host | Hostname of the container registry from which to pull model image containers from | N/A | |
| model.registry.port | The port that registry is listening on | 443 | |
| modzy.registry.credentials | Credentials necessary for pulling images from the specified registry | N/A | |
| modzy.registry.skipTLSVerify | Whether or not to skip the TLS verification step when connecting to the model registry | False | |
| storage.driver | Modzy Edge will store job data and results using the specified storage driver. Avaialble options are: S3 Local Filesystem * Remote DB [Only Local is implemented] | N/A -- we should make it filesystem tho? | True |
| storage.filesystem.path | The directory on the local filesystem to save job data | ~/.modzy or /opt/modzy | True if your storage.driver = filesystem |
| storage.s3.region | The region in which your S3 bucket resides | N/A | True if your storage.driver = S3 |
| storage.s3.bucket-name | The name of the S3 bucket used to save job data | N/A | True if your storage.driver = S3 |
| storage.s3.access-key | The access key for your S3 bucket | N/A | True if your storage.driver = S3 |
| storage.s3.access-key-secret | The secret access key for your bucket | N/A | True if your storage.driver = S3 |
| storage.s3.path | The path in the S3 bucket where job data will be saved | N/A | True if your storage.driver = S3 |
Models
You'll need to select at least one model from your Model Library to associate with your Device Group. Nothing changes when discussing models in the context of a Device Group -- these are the same models available to you in your Modzy Enterprise installation and work exactly the same.
When Modzy Edge boots up on your device, it will inspect the local container image registry on that device for the model images . If the model image(s) is found, then Modzy Edge will launch the container on startup. In the case that the images are not present in the local registry, then Modzy Edge will connect to the specified container registry and pull the necessary image before starting the model container.
When it comes to which model is deployed, it's an all or nothing situation. Modzy Edge will run all models associated with the Device Group on the host device, so it's important to add only the models you'll need to run on the devices associated with the Device Group.
The Registration Token is exchanged between Modzy Edge, and the enterprise installation of Modzy that launched it. This handshake establishes the validity of the deployment, ensuring that the deployment on a device came from a valid Device Group from a valid enterprise installation of Modzy.