Altair SmartWorks Analytics


Setting Up a Seldon Connection

A Seldon connection under the Pandas Execution Engine is necessary to use the MLOps Deployment functionality. This connection to Seldon is what is referred to as the Deployment Profile in MLOps.

Note the following details when working with Seldon:

  • Versions of Seldon Core >= 1.4.0 are supported by the application.

  • For the analytics to work, the Seldon Core analytics module should be installed on the cluster in the same namespace in which the Seldon Core Operator is installed and where deployments will be applied.

  • For Autoscaling to work, the Kubernetes Metric Server should be installed on the cluster.

  • While any API Gateway may be used, Seldon officially supports Ambassador and Istio.


  1. Open a Pandas-based Execution Profile and then click the Internal Connections tab.

  2. In the List of Connections page, click the Add New Internal Connection button.

  3. The New Connection page displays.


  4. Specify the connection type Application and the application type Seldon.

  5. Provide a connection name (required) and description (optional).

  6. Specify the following properties:

  7. Property


    Cluster Manager

    Currently, only AWS is supported.

    Cluster Name

    The name of the cluster


    The AWS region in which the cluster is located

    Access Key

    The AWS Access Key

    Secret Key

    The AWS Secret Key

    Gateway Endpoint

    This can be any gateway URL that routes the requests to Seldon

    Seldon Namespace

    The namespace where the Seldon Core Operator is installed and models will be deployed

    Has GPU Nodes

    Check this box if your Kubernetes namespace contains GPU nodes

    Additional GPU Configurations

    Specify your Kubernetes nodeSelector and tolerations in yaml format to target the specific nodes that have GPUs


  8. Click Save when you are finished.

  9. Your settings are saved, and the new Seldon connection is added to the Internal Connections page.