The scenario describes a compute cluster that is "fully-utilized" and cannot meet the inferencing demand, indicating that its maximum capacity has been reached. The most direct, efficient, and least disruptive way to increase the total processing capacity of a compute cluster like Azure Kubernetes Service (AKS) is through horizontal scaling, which involves adding more compute nodes. This action allows the service to handle more concurrent requests by distributing the load across a larger number of machines. This approach directly addresses the performance bottleneck with minimal administrative effort and without requiring service redeployment, thus ensuring minimal downtime.

A. Increasing the minimum node count only affects the baseline number of nodes for autoscaling; it does not raise the maximum capacity, which is the current limitation.

B. Creating a new cluster and redeploying is a complex migration process (blue-green deployment) that requires significant administrative effort and planning, contradicting the requirements.

C. Increasing the VM size (vertical scaling) is more disruptive than adding nodes, often requiring the creation of a new node pool and draining the old one, which is not a minimal effort task.

1. Microsoft Azure Documentation

"Troubleshoot model deployment": In the section on autoscaling

the documentation explains that when a service fails because the cluster is at maximum capacity

the solution is to increase the capacity. It states

"To change the maximum number of nodes in the cluster

you need to update the compute target." This directly supports increasing the node count to handle higher loads.

Source: Microsoft Corporation. (2023). Troubleshoot model deployment with Azure Machine Learning. Microsoft Docs. Retrieved from https://docs.microsoft.com/en-us/azure/machine-learning/how-to-troubleshoot-deployment#autoscaling (Refer to the section on Autoscaling).

2. Microsoft Azure Documentation

"Scale cluster in Azure Kubernetes Service (AKS)": This document details the procedures for scaling an AKS cluster. It highlights that scaling the node count is the primary method for adjusting compute capacity to meet application demands. The process can be done manually with a single command (az aks scale)

which exemplifies minimal administrative effort and is designed to be non-disruptive to running applications.

Source: Microsoft Corporation. (2023). Scale the node count in an Azure Kubernetes Service (AKS) cluster. Microsoft Docs. Retrieved from https://docs.microsoft.com/en-us/azure/aks/scale-cluster (Refer to the "Manually scale cluster nodes" section).

3. Microsoft Azure Documentation

"Deploy a model to an Azure Kubernetes Service cluster": This guide explains how to configure an AKS cluster for machine learning workloads

including setting autoscaling parameters. It notes that if autoscaling is not enabled

the cluster operates with a fixed number of nodes

which can be changed manually. If autoscaling is enabled but the maximum is reached

the solution is to increase the autoscalemaxnodes parameter when updating the compute target. Both actions are equivalent to increasing the node count.

Source: Microsoft Corporation. (2023). Deploy a model to an Azure Kubernetes Service cluster. Microsoft Docs. Retrieved from https://docs.microsoft.com/en-us/azure/machine-learning/how-to-deploy-azure-kubernetes-service (Refer to the "Autoscaling" section).