Data Storage
Azure Storage is a cloud-based service that provides highly available, scalable, and secure storage for a wide variety of data objects.
Compute
The cloud computing service from Microsoft Azure gives users an infrastructure that lets them run applications and change the amount of capacity based on their needs.
There are four primary compute types available in Azure ML:
- Compute instances
- Compute clusters
- Inference clusters
- Attached compute
Compute Instances
Azure Compute instances give you fully managed virtual machines that already have most of the important frameworks and libraries for doing machine learning and data science experiments. This saves you time because you don’t have to set up your local computer for ML experiments.
Compute Clusters
When compared to compute instances, compute clusters can have more than one compute node. We can customize the hardware specs of these compute nodes to suit our needs. A compute cluster may need to be set up for hyperparameter tuning, GPU-based complex computations, and running multiple machine learning models in parallel. You will need a compute cluster to run computations when using Azure ML for automated machine learning experiments.
Inference Clusters
Azure Machine Learning web services can be set up as REST endpoints in one of two ways:
| 1. | Implement ACI (Azure Container Instances): For smaller, more localized deployments of the REST web service, ACI is the way to go. | |
| 2. | Provision AKS (Azure Kubernetes Service): AKS is the way to go for larger, more enterprise-wide deployments. The AKS cluster can be adjusted to your specifications via the Azure portal and Azure ML. |
Attached Compute
Azure Machine Learning isn’t just about computing on compute clusters. You can use Azure Databricks, Data Lake Analytics, HDInsight, or a VM you already have as a computing resource for your workspace. Keep in mind that Azure Machine Learning only works with Ubuntu virtual machines. Azure Machine Learning will not manage these compute targets. Because of this, you might need to take some extra steps to make sure they work with your experiments.
Services
Azure’s services follow a pay-as-you-go business model and are made available to customers over the Internet. Servers, storage, databases, networking, software, analytics, and a variety of services are available for the AI and ML category, including Azure Cognitive Services, Azure Machine Learning service, Azure Cognitive Search, and Azure Bot Services, to name a few.