Working with Multiple Kubernetes Compute Server Connection

Kyvos supports multiple Compute Server Clusters, each associated with a dedicated K8S cluster or user pool.

You can define multiple compute server cluster connections, and while processing a semantic model job, you can select the required connection while adding a semantic model job.

If you want complete resource isolation, you can assign different compute server cluster connections for each semantic model process. This ensures that workloads remain isolated, performance is predictable, and a high-priority semantic model can run independently.

To add multiple Kubernetes Compute Master connections, perform the following steps.

From the Toolbox, click Connections.
From the Actions menu ( ⋮ ) click Add Connection. The Connection page is displayed.
After you have entered the parameters described in the table below, click the Test button in the top left to validate the connection settings.
If the connection is valid, click the Save button.
To refresh connections, click the Actions menu (⋮) at the top of the Connections column and select Refresh.

Parameter/Field	Comments/Description
Name	Enter a unique name for the connection.
Category	Select the Compute Server option.
Provider	Select the Kubernetes option.
Kubernetes Cluster	Select the Kubernetes cluster from the list.
Agent Pool	Select the AKS agent pool for the new connection from the list.
User Pool	Select the AKS User pool from the list where all Kyvos workloads will run.
User Pool Node Memory	The number of user pool node memory is prepopulated (read-only mode).
User Pool Node Core	The number of user pool node cores is prepopulated (read-only mode).
User Pool Maximum Node	It defines the limit on the number of nodes available for the Kyvos workload. You can update it as needed.
Resource Group	This is an AKS cluster resource group in read-only mode. It is shared for all Kubernetes-based connections, including the default read-only connection.
Namespace	Kyvos workload-specific namespace where all the Kyvos pods will run. This is read-only mode and is shared for all Kubernetes-based connections, including the default connection, which is also read-only.
Taint	Enter the user pool taint first name.
Compute Server Memory Config Mode	Select one of the following memory mode configurations for the compute server process from the list: Auto: Kyvos will automatically calculate both heap and tabular store memory based on the total memory of the user pool node. Manual: The user can provide heap and tabular store memory within the limit of the total memory of the user pool node
Concurrent Task Mode	Select one of the following concurrent tasks that can run on the compute server: Auto: Kyvos will automatically calculate the user pool node capacity based on the total memory and total cores of the user pool, using the predefined formula. Manual: The user can provide a concurrent tasks limit for the compute server within the total cores limit of the user pool node.
Compute Server Startup Count	The user can specify the number of compute server instances to create when launching the semantic model process job.
Is Data Process	The Is Data Process checkbox is selected by default. Defines that the connection can be used for the semantic model process. The connection will be listed for the semantic model process during job scheduling.
Properties	Click Properties to view or set properties.