Azure Machine Learning Workspace / Azure AI Foundry - Complete Deep-Dive Reference
Authoritative reference for every aspect of Azure Machine Learning Workspace (Azure AI Foundry) including architecture, networking, private endpoints, compute clusters, endpoint deployment, managed identities, ACR integration, storage accounts, all CLI and PowerShell commands, log reading, debugging, and Terraform integration.
1. ARCHITECTURE AND CORE CONCEPTS
Workspace Resource Hierarchy
Azure Subscription
└── Resource Group
├── Azure ML Workspace (Microsoft.MachineLearningServices/workspaces)
│ ├── Dependent Resources (auto-created or BYO)
│ │ ├── Azure Storage Account (default datastore)
│ │ ├── Azure Key Vault (secrets, connection strings)
│ │ ├── Azure Application Insights (telemetry)
│ │ └── Azure Container Registry (Docker images for environments)
│ ├── Compute Targets
│ │ ├── Compute Instances (dev/test VMs)
│ │ ├── Compute Clusters (AmlCompute - training)
│ │ ├── Serverless Compute (on-demand)
│ │ ├── Kubernetes Compute (AKS / Arc-enabled)
│ │ └── Attached Compute (Databricks, HDInsight, VMs)
│ ├── Data Assets (versioned references to data)
│ ├── Datastores (connections to storage)
│ ├── Environments (Docker + conda specs)
│ ├── Models (registered trained models)
│ ├── Endpoints
│ │ ├── Managed Online Endpoints (real-time)
│ │ ├── Kubernetes Online Endpoints (BYO infra)
│ │ ├── Batch Endpoints (large-scale scoring)
│ │ └── Serverless Endpoints (MaaS - pay-per-token)
│ ├── Jobs (training runs, pipelines, sweeps)
│ ├── Components (reusable pipeline steps)
│ ├── Schedules (recurring job triggers)
│ └── Registries (cross-workspace sharing)
└── AI Foundry Hub (kind=hub) + Projects (kind=project)
AI Foundry Hub/Project vs Classic Workspace
| Feature |
Classic Workspace (kind=Default) |
AI Foundry Hub + Project |
| Portal |
ml.azure.com |
ai.azure.com |
| Scope |
Single workspace |
Hub shares infra across projects |
| Networking |
Per-workspace |
Hub-level (shared across projects) |
| Identity |
Per-workspace |
Hub-level identity, project inherits |
| Model catalog |
Yes |
Yes, plus additional Foundry models |
| Prompt flow |
Yes |
Yes |
| AI agents |
Limited |
Full AI Agent Service |
| Use case |
Classical ML, custom training |
GenAI, LLM apps, AI agents |
Workspace Creation - All Methods
CLI:
# Install/upgrade ML extension
az extension add --name ml --upgrade
# Create resource group
az group create --name ml-rg --location eastus
# Create workspace with all dependencies auto-created
az ml workspace create \
--name my-ml-workspace \
--resource-group ml-rg \
--location eastus
# Create workspace with explicit dependencies
az ml workspace create \
--name my-ml-workspace \
--resource-group ml-rg \
--location eastus \
--storage-account /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.Storage/storageAccounts/mlstorage \
--key-vault /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.KeyVault/vaults/mlkeyvault \
--app-insights /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.Insights/components/mlinsights \
--container-registry /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.ContainerRegistry/registries/mlacr \
--public-network-access Disabled \
--managed-network AllowInternetOutbound \
--image-build-compute cpu-build-cluster \
--enable-data-isolation true \
--tags Environment=Production Team=DataScience
# Create AI Foundry Hub
az ml workspace create \
--name my-ai-hub \
--resource-group ml-rg \
--location eastus \
--kind hub \
--storage-account aihubstorage \
--key-vault aihubkeyvault
# Create AI Foundry Project within Hub
az ml workspace create \
--name my-ai-project \
--resource-group ml-rg \
--location eastus \
--kind project \
--hub-id /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.MachineLearningServices/workspaces/my-ai-hub
# Show workspace details
az ml workspace show \
--name my-ml-workspace \
--resource-group ml-rg
# List all workspaces
az ml workspace list \
--resource-group ml-rg \
--output table
# Update workspace
az ml workspace update \
--name my-ml-workspace \
--resource-group ml-rg \
--description "Updated workspace" \
--public-network-access Disabled
# Delete workspace
az ml workspace delete \
--name my-ml-workspace \
--resource-group ml-rg \
--permanently-delete --all-resources
# Diagnose workspace configuration
az ml workspace diagnose \
--name my-ml-workspace \
--resource-group ml-rg
PowerShell (Az.MachineLearningServices):
# Install the module
Install-Module -Name Az.MachineLearningServices -Scope CurrentUser -Repository PSGallery -Force
# Create workspace
New-AzMLWorkspace `
-Name "my-ml-workspace" `
-ResourceGroupName "ml-rg" `
-Location "eastus" `
-StorageAccountId "/subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.Storage/storageAccounts/mlstorage" `
-KeyVaultId "/subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.KeyVault/vaults/mlkeyvault" `
-ApplicationInsightId "/subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.Insights/components/mlinsights" `
-IdentityType "SystemAssigned" `
-PublicNetworkAccess "Disabled"
# Get workspace
Get-AzMLWorkspace -Name "my-ml-workspace" -ResourceGroupName "ml-rg"
# List workspaces
Get-AzMLWorkspace -ResourceGroupName "ml-rg"
# Update workspace
Update-AzMLWorkspace `
-Name "my-ml-workspace" `
-ResourceGroupName "ml-rg" `
-Description "Updated workspace" `
-Tag @{Environment="Production"}
# Remove workspace
Remove-AzMLWorkspace -Name "my-ml-workspace" -ResourceGroupName "ml-rg"
# Diagnose workspace
Invoke-AzMLWorkspaceDiagnose -Name "my-ml-workspace" -ResourceGroupName "ml-rg"
2. NETWORKING
Azure ML supports three managed network isolation modes (Disabled, AllowInternetOutbound, AllowOnlyApprovedOutbound) with the managed VNet approach recommended for production. Private endpoints provide inbound connectivity, and outbound rules control egress from compute resources.
Key DNS Zones
| Service |
Private DNS Zone |
| ML Workspace API |
privatelink.api.azureml.ms |
| ML Notebooks |
privatelink.notebooks.azure.net |
| Storage Blob |
privatelink.blob.core.windows.net |
| Storage File |
privatelink.file.core.windows.net |
| Key Vault |
privatelink.vaultcore.azure.net |
| Container Registry |
privatelink.azurecr.io |
| Application Insights |
privatelink.monitor.azure.com |
Key Service Tags
| Service Tag |
Purpose |
| AzureMachineLearning |
ML workspace management (inbound 44224, outbound 443) |
| BatchNodeManagement |
Compute cluster management (inbound 29876-29877) |
| Storage |
Access to Azure Storage (outbound 443) |
| AzureActiveDirectory |
Authentication (outbound 443) |
For full VNet configuration, private endpoint setup, NSG rules, and outbound rule management, see references/networking.md.
3. COMPUTE
Azure ML offers multiple compute targets: Compute Instances for dev/test, AmlCompute Clusters for scalable training, Serverless Compute for on-demand jobs without cluster management, and Kubernetes Compute for BYO infrastructure scenarios.
GPU VM SKU Quick Reference
| VM Series |
GPU |
GPU Memory |
Use Case |
| Standard_NC24ads_A100_v4 |
1x A100 |
80 GB |
Training, fine-tuning |
| Standard_ND96amsr_A100_v4 |
8x A100 80GB |
640 GB |
Large model training |
| Standard_ND_H100_v5 |
8x H100 |
640 GB |
GenAI, LLM training |
| Standard_ND_H200_v5 |
8x H200 |
1120 GB |
Latest: 2x perf vs H100 |
| Standard_NCads_H100_v5 |
1x H100 NVL |
94 GB |
Inference, fine-tuning |
| Standard_NC4as_T4_v3 |
1x T4 |
16 GB |
Budget inference |
For the complete GPU SKU table, compute instance/cluster CLI reference, serverless compute, Kubernetes attach, and debugging commands, see references/compute.md.
4. ENDPOINT DEPLOYMENT
Azure ML supports four endpoint types: Managed Online Endpoints (recommended for real-time inference with blue-green deployments), Batch Endpoints (large-scale scoring on compute clusters), Kubernetes Online Endpoints (BYO AKS/Arc infrastructure), and Serverless Endpoints (pay-per-token Model-as-a-Service).
Endpoint Types Quick Reference
| Type |
Use Case |
Auth Modes |
Scaling |
| Managed Online |
Real-time inference |
key, aml_token |
Per-deployment instance count |
| Batch |
Large-scale scoring |
managed identity |
Compute cluster auto-scale |
| Kubernetes Online |
BYO infra real-time |
key, aml_token |
K8s pod scaling |
| Serverless (MaaS) |
Pay-per-token LLM |
key |
Automatic |
For full endpoint creation, deployment, traffic splitting, log retrieval, and batch invocation commands, see references/endpoints.md.
5-7. IDENTITIES, ACR, AND STORAGE
Managed identities (system-assigned or user-assigned) control access between workspace, compute, endpoints, and dependent resources. ACR stores Docker images for environments and model serving, requiring Premium SKU for private endpoints and an image-build-compute cluster when behind a VNet. Storage accounts serve as the default datastore for blobs, file shares, job outputs, and MLflow artifacts.
Identity Types
| Identity Type |
Use Case |
| System-Assigned (workspace) |
Default workspace operations, auto-lifecycle |
| User-Assigned (workspace) |
CMK encryption, cross-resource sharing |
| System-Assigned (compute) |
Per-cluster storage/ACR access |
| User-Assigned (compute) |
Fine-grained, reusable access control |
Key RBAC Roles
| Role |
Description |
| AzureML Data Scientist |
Run jobs, manage compute, deploy models |
| AzureML Compute Operator |
Create/manage compute resources |
| Azure AI Developer |
AI Foundry project development |
| Azure AI Inference Deployment Operator |
Deploy models to endpoints |
For full identity configuration, role assignment commands, ACR integration, private ACR setup, datastore registration, and storage account details, see references/identities-acr-storage.md.
8-9. CLI AND POWERSHELL
The az ml CLI extension provides comprehensive workspace management through 20+ command groups covering workspaces, compute, jobs, models, endpoints, environments, data, datastores, components, schedules, registries, and connections. The Az.MachineLearningServices PowerShell module offers equivalent functionality for Windows-native automation.
Key az ml Command Groups
| Command Group |
Purpose |
az ml workspace |
Manage workspaces (create, diagnose, provision-network, outbound-rule) |
az ml compute |
Manage compute (create, start, stop, connect-ssh, attach) |
az ml job |
Manage jobs (create, stream, cancel, download) |
az ml online-endpoint |
Manage online endpoints (create, invoke, get-credentials) |
az ml online-deployment |
Manage deployments (create, get-logs, traffic) |
az ml batch-endpoint |
Manage batch endpoints (create, invoke, list-jobs) |
az ml serverless-endpoint |
Manage serverless endpoints (create, get-credentials) |
For the complete command reference, job management deep-dive, schedule management, and full PowerShell cmdlet reference, see references/cli-powershell.md.
10. TERRAFORM INTEGRATION
Azure ML workspaces can be fully provisioned with Terraform using the azurerm provider. A production setup includes the workspace, VNet/subnets, NSG, storage account, key vault, ACR, Application Insights, private endpoints, DNS zones, compute clusters, and RBAC role assignments.
Key Terraform Resources
| Resource |
Purpose |
azurerm_machine_learning_workspace |
ML workspace (Default, Hub, Project) |
azurerm_machine_learning_compute_cluster |
AmlCompute training clusters |
azurerm_machine_learning_compute_instance |
Dev/test compute instances |
azurerm_machine_learning_workspace_network_outbound_rule_* |
Managed network outbound rules |
For the full production-ready Terraform configuration (providers, networking, storage, key vault, ACR, workspace, compute, role assignments, and outputs), see references/terraform.md.
11. TROUBLESHOOTING AND DEBUGGING
Azure ML provides multiple debugging surfaces: real-time job log streaming, deployment container logs (inference-server and storage-initializer), compute instance SSH access for system-level diagnostics, Log Analytics queries for historical analysis, and the az ml workspace diagnose command for configuration validation.
Common Error Categories
| Category |
Common Errors |
| Compute |
QuotaExceeded, AllocationFailed, disk full, GPU not detected |
| Endpoints |
ScoringError, HealthCheckFailure, ImageBuildFailed, 429/503 errors |
| Networking |
DNS resolution failure, connection timeout, storage/ACR access denied |
| Jobs |
EnvironmentBuildError, OutOfMemoryError, NCCL timeout, blob not found |
For full error reference tables, log locations, Log Analytics queries, endpoint metrics monitoring, workspace diagnostics, and the secure workspace setup checklist, see references/troubleshooting.md.
Additional Resources
Detailed reference files for each topic area:
- references/networking.md -- VNet, private endpoints, DNS zones, NSG rules, service tags
- references/compute.md -- GPU SKUs, compute instances, clusters, serverless, Kubernetes
- references/endpoints.md -- Managed online, batch, Kubernetes, and serverless endpoints
- references/identities-acr-storage.md -- Managed identities, ACR integration, storage accounts
- references/cli-powershell.md -- Complete az ml CLI and PowerShell command reference
- references/terraform.md -- Full production-ready Terraform configuration
- references/troubleshooting.md -- Log reading, debugging, error tables, setup checklist
External Documentation
1---
2name: azure-ml-foundry-workspace
3description: |
4 Azure Machine Learning Workspace and Azure AI Foundry deep dive.
5 PROACTIVELY activate for: (1) creating and configuring Azure ML workspaces, (2) Azure AI Foundry hubs and projects, (3) ML workspace networking (managed VNet, private endpoints, DNS), (4) ML compute clusters and compute instances, (5) managed online endpoints, batch endpoints, Kubernetes endpoints, (6) managed identities for ML resources, (7) ACR integration for custom environments, (8) storage account configuration, (9) az ml CLI v2 commands, (10) PowerShell Az.MachineLearningServices, (11) reading ML compute and deployment logs, (12) GPU SKU selection (ND/NC series, H100/H200/A100).
6 Provides: workspace setup playbook, network-isolation patterns, endpoint deployment templates, az ml CLI cheat sheet, and log diagnosis workflow.
7metadata:
8 source: https://skillmd.com/skills/josiahsiegel/azure-ml-foundry-workspace
9---
10
11# Azure Machine Learning Workspace / Azure AI Foundry - Complete Deep-Dive Reference
12
13Authoritative reference for every aspect of Azure Machine Learning Workspace (Azure AI Foundry) including architecture, networking, private endpoints, compute clusters, endpoint deployment, managed identities, ACR integration, storage accounts, all CLI and PowerShell commands, log reading, debugging, and Terraform integration.
14
15---
16
17## 1. ARCHITECTURE AND CORE CONCEPTS
18
19### Workspace Resource Hierarchy
20
21```text
22Azure Subscription
23 └── Resource Group
24 ├── Azure ML Workspace (Microsoft.MachineLearningServices/workspaces)
25 │ ├── Dependent Resources (auto-created or BYO)
26 │ │ ├── Azure Storage Account (default datastore)
27 │ │ ├── Azure Key Vault (secrets, connection strings)
28 │ │ ├── Azure Application Insights (telemetry)
29 │ │ └── Azure Container Registry (Docker images for environments)
30 │ ├── Compute Targets
31 │ │ ├── Compute Instances (dev/test VMs)
32 │ │ ├── Compute Clusters (AmlCompute - training)
33 │ │ ├── Serverless Compute (on-demand)
34 │ │ ├── Kubernetes Compute (AKS / Arc-enabled)
35 │ │ └── Attached Compute (Databricks, HDInsight, VMs)
36 │ ├── Data Assets (versioned references to data)
37 │ ├── Datastores (connections to storage)
38 │ ├── Environments (Docker + conda specs)
39 │ ├── Models (registered trained models)
40 │ ├── Endpoints
41 │ │ ├── Managed Online Endpoints (real-time)
42 │ │ ├── Kubernetes Online Endpoints (BYO infra)
43 │ │ ├── Batch Endpoints (large-scale scoring)
44 │ │ └── Serverless Endpoints (MaaS - pay-per-token)
45 │ ├── Jobs (training runs, pipelines, sweeps)
46 │ ├── Components (reusable pipeline steps)
47 │ ├── Schedules (recurring job triggers)
48 │ └── Registries (cross-workspace sharing)
49 └── AI Foundry Hub (kind=hub) + Projects (kind=project)
50```
51
52### AI Foundry Hub/Project vs Classic Workspace
53
54| Feature | Classic Workspace (kind=Default) | AI Foundry Hub + Project |
55|---------|----------------------------------|--------------------------|
56| Portal | ml.azure.com | ai.azure.com |
57| Scope | Single workspace | Hub shares infra across projects |
58| Networking | Per-workspace | Hub-level (shared across projects) |
59| Identity | Per-workspace | Hub-level identity, project inherits |
60| Model catalog | Yes | Yes, plus additional Foundry models |
61| Prompt flow | Yes | Yes |
62| AI agents | Limited | Full AI Agent Service |
63| Use case | Classical ML, custom training | GenAI, LLM apps, AI agents |
64
65### Workspace Creation - All Methods
66
67**CLI:**
68```bash
69# Install/upgrade ML extension
70az extension add --name ml --upgrade
71
72# Create resource group
73az group create --name ml-rg --location eastus
74
75# Create workspace with all dependencies auto-created
76az ml workspace create \
77 --name my-ml-workspace \
78 --resource-group ml-rg \
79 --location eastus
80
81# Create workspace with explicit dependencies
82az ml workspace create \
83 --name my-ml-workspace \
84 --resource-group ml-rg \
85 --location eastus \
86 --storage-account /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.Storage/storageAccounts/mlstorage \
87 --key-vault /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.KeyVault/vaults/mlkeyvault \
88 --app-insights /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.Insights/components/mlinsights \
89 --container-registry /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.ContainerRegistry/registries/mlacr \
90 --public-network-access Disabled \
91 --managed-network AllowInternetOutbound \
92 --image-build-compute cpu-build-cluster \
93 --enable-data-isolation true \
94 --tags Environment=Production Team=DataScience
95
96# Create AI Foundry Hub
97az ml workspace create \
98 --name my-ai-hub \
99 --resource-group ml-rg \
100 --location eastus \
101 --kind hub \
102 --storage-account aihubstorage \
103 --key-vault aihubkeyvault
104
105# Create AI Foundry Project within Hub
106az ml workspace create \
107 --name my-ai-project \
108 --resource-group ml-rg \
109 --location eastus \
110 --kind project \
111 --hub-id /subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.MachineLearningServices/workspaces/my-ai-hub
112
113# Show workspace details
114az ml workspace show \
115 --name my-ml-workspace \
116 --resource-group ml-rg
117
118# List all workspaces
119az ml workspace list \
120 --resource-group ml-rg \
121 --output table
122
123# Update workspace
124az ml workspace update \
125 --name my-ml-workspace \
126 --resource-group ml-rg \
127 --description "Updated workspace" \
128 --public-network-access Disabled
129
130# Delete workspace
131az ml workspace delete \
132 --name my-ml-workspace \
133 --resource-group ml-rg \
134 --permanently-delete --all-resources
135
136# Diagnose workspace configuration
137az ml workspace diagnose \
138 --name my-ml-workspace \
139 --resource-group ml-rg
140```
141
142**PowerShell (Az.MachineLearningServices):**
143```powershell
144# Install the module
145Install-Module -Name Az.MachineLearningServices -Scope CurrentUser -Repository PSGallery -Force
146
147# Create workspace
148New-AzMLWorkspace `
149 -Name "my-ml-workspace" `
150 -ResourceGroupName "ml-rg" `
151 -Location "eastus" `
152 -StorageAccountId "/subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.Storage/storageAccounts/mlstorage" `
153 -KeyVaultId "/subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.KeyVault/vaults/mlkeyvault" `
154 -ApplicationInsightId "/subscriptions/<sub>/resourceGroups/ml-rg/providers/Microsoft.Insights/components/mlinsights" `
155 -IdentityType "SystemAssigned" `
156 -PublicNetworkAccess "Disabled"
157
158# Get workspace
159Get-AzMLWorkspace -Name "my-ml-workspace" -ResourceGroupName "ml-rg"
160
161# List workspaces
162Get-AzMLWorkspace -ResourceGroupName "ml-rg"
163
164# Update workspace
165Update-AzMLWorkspace `
166 -Name "my-ml-workspace" `
167 -ResourceGroupName "ml-rg" `
168 -Description "Updated workspace" `
169 -Tag @{Environment="Production"}
170
171# Remove workspace
172Remove-AzMLWorkspace -Name "my-ml-workspace" -ResourceGroupName "ml-rg"
173
174# Diagnose workspace
175Invoke-AzMLWorkspaceDiagnose -Name "my-ml-workspace" -ResourceGroupName "ml-rg"
176```
177
178---
179
180## 2. NETWORKING
181
182Azure ML supports three managed network isolation modes (Disabled, AllowInternetOutbound, AllowOnlyApprovedOutbound) with the managed VNet approach recommended for production. Private endpoints provide inbound connectivity, and outbound rules control egress from compute resources.
183
184### Key DNS Zones
185
186| Service | Private DNS Zone |
187|---------|-----------------|
188| ML Workspace API | privatelink.api.azureml.ms |
189| ML Notebooks | privatelink.notebooks.azure.net |
190| Storage Blob | privatelink.blob.core.windows.net |
191| Storage File | privatelink.file.core.windows.net |
192| Key Vault | privatelink.vaultcore.azure.net |
193| Container Registry | privatelink.azurecr.io |
194| Application Insights | privatelink.monitor.azure.com |
195
196### Key Service Tags
197
198| Service Tag | Purpose |
199|------------|---------|
200| AzureMachineLearning | ML workspace management (inbound 44224, outbound 443) |
201| BatchNodeManagement | Compute cluster management (inbound 29876-29877) |
202| Storage | Access to Azure Storage (outbound 443) |
203| AzureActiveDirectory | Authentication (outbound 443) |
204
205For full VNet configuration, private endpoint setup, NSG rules, and outbound rule management, see **[references/networking.md](references/networking.md)**.
206
207---
208
209## 3. COMPUTE
210
211Azure ML offers multiple compute targets: Compute Instances for dev/test, AmlCompute Clusters for scalable training, Serverless Compute for on-demand jobs without cluster management, and Kubernetes Compute for BYO infrastructure scenarios.
212
213### GPU VM SKU Quick Reference
214
215| VM Series | GPU | GPU Memory | Use Case |
216|-----------|-----|-----------|----------|
217| Standard_NC24ads_A100_v4 | 1x A100 | 80 GB | Training, fine-tuning |
218| Standard_ND96amsr_A100_v4 | 8x A100 80GB | 640 GB | Large model training |
219| Standard_ND_H100_v5 | 8x H100 | 640 GB | GenAI, LLM training |
220| Standard_ND_H200_v5 | 8x H200 | 1120 GB | Latest: 2x perf vs H100 |
221| Standard_NCads_H100_v5 | 1x H100 NVL | 94 GB | Inference, fine-tuning |
222| Standard_NC4as_T4_v3 | 1x T4 | 16 GB | Budget inference |
223
224For the complete GPU SKU table, compute instance/cluster CLI reference, serverless compute, Kubernetes attach, and debugging commands, see **[references/compute.md](references/compute.md)**.
225
226---
227
228## 4. ENDPOINT DEPLOYMENT
229
230Azure ML supports four endpoint types: Managed Online Endpoints (recommended for real-time inference with blue-green deployments), Batch Endpoints (large-scale scoring on compute clusters), Kubernetes Online Endpoints (BYO AKS/Arc infrastructure), and Serverless Endpoints (pay-per-token Model-as-a-Service).
231
232### Endpoint Types Quick Reference
233
234| Type | Use Case | Auth Modes | Scaling |
235|------|----------|------------|---------|
236| Managed Online | Real-time inference | key, aml_token | Per-deployment instance count |
237| Batch | Large-scale scoring | managed identity | Compute cluster auto-scale |
238| Kubernetes Online | BYO infra real-time | key, aml_token | K8s pod scaling |
239| Serverless (MaaS) | Pay-per-token LLM | key | Automatic |
240
241For full endpoint creation, deployment, traffic splitting, log retrieval, and batch invocation commands, see **[references/endpoints.md](references/endpoints.md)**.
242
243---
244
245## 5-7. IDENTITIES, ACR, AND STORAGE
246
247Managed identities (system-assigned or user-assigned) control access between workspace, compute, endpoints, and dependent resources. ACR stores Docker images for environments and model serving, requiring Premium SKU for private endpoints and an image-build-compute cluster when behind a VNet. Storage accounts serve as the default datastore for blobs, file shares, job outputs, and MLflow artifacts.
248
249### Identity Types
250
251| Identity Type | Use Case |
252|--------------|----------|
253| System-Assigned (workspace) | Default workspace operations, auto-lifecycle |
254| User-Assigned (workspace) | CMK encryption, cross-resource sharing |
255| System-Assigned (compute) | Per-cluster storage/ACR access |
256| User-Assigned (compute) | Fine-grained, reusable access control |
257
258### Key RBAC Roles
259
260| Role | Description |
261|------|-------------|
262| AzureML Data Scientist | Run jobs, manage compute, deploy models |
263| AzureML Compute Operator | Create/manage compute resources |
264| Azure AI Developer | AI Foundry project development |
265| Azure AI Inference Deployment Operator | Deploy models to endpoints |
266
267For full identity configuration, role assignment commands, ACR integration, private ACR setup, datastore registration, and storage account details, see **[references/identities-acr-storage.md](references/identities-acr-storage.md)**.
268
269---
270
271## 8-9. CLI AND POWERSHELL
272
273The `az ml` CLI extension provides comprehensive workspace management through 20+ command groups covering workspaces, compute, jobs, models, endpoints, environments, data, datastores, components, schedules, registries, and connections. The `Az.MachineLearningServices` PowerShell module offers equivalent functionality for Windows-native automation.
274
275### Key az ml Command Groups
276
277| Command Group | Purpose |
278|--------------|---------|
279| `az ml workspace` | Manage workspaces (create, diagnose, provision-network, outbound-rule) |
280| `az ml compute` | Manage compute (create, start, stop, connect-ssh, attach) |
281| `az ml job` | Manage jobs (create, stream, cancel, download) |
282| `az ml online-endpoint` | Manage online endpoints (create, invoke, get-credentials) |
283| `az ml online-deployment` | Manage deployments (create, get-logs, traffic) |
284| `az ml batch-endpoint` | Manage batch endpoints (create, invoke, list-jobs) |
285| `az ml serverless-endpoint` | Manage serverless endpoints (create, get-credentials) |
286
287For the complete command reference, job management deep-dive, schedule management, and full PowerShell cmdlet reference, see **[references/cli-powershell.md](references/cli-powershell.md)**.
288
289---
290
291## 10. TERRAFORM INTEGRATION
292
293Azure ML workspaces can be fully provisioned with Terraform using the `azurerm` provider. A production setup includes the workspace, VNet/subnets, NSG, storage account, key vault, ACR, Application Insights, private endpoints, DNS zones, compute clusters, and RBAC role assignments.
294
295### Key Terraform Resources
296
297| Resource | Purpose |
298|----------|---------|
299| `azurerm_machine_learning_workspace` | ML workspace (Default, Hub, Project) |
300| `azurerm_machine_learning_compute_cluster` | AmlCompute training clusters |
301| `azurerm_machine_learning_compute_instance` | Dev/test compute instances |
302| `azurerm_machine_learning_workspace_network_outbound_rule_*` | Managed network outbound rules |
303
304For the full production-ready Terraform configuration (providers, networking, storage, key vault, ACR, workspace, compute, role assignments, and outputs), see **[references/terraform.md](references/terraform.md)**.
305
306---
307
308## 11. TROUBLESHOOTING AND DEBUGGING
309
310Azure ML provides multiple debugging surfaces: real-time job log streaming, deployment container logs (inference-server and storage-initializer), compute instance SSH access for system-level diagnostics, Log Analytics queries for historical analysis, and the `az ml workspace diagnose` command for configuration validation.
311
312### Common Error Categories
313
314| Category | Common Errors |
315|----------|--------------|
316| Compute | QuotaExceeded, AllocationFailed, disk full, GPU not detected |
317| Endpoints | ScoringError, HealthCheckFailure, ImageBuildFailed, 429/503 errors |
318| Networking | DNS resolution failure, connection timeout, storage/ACR access denied |
319| Jobs | EnvironmentBuildError, OutOfMemoryError, NCCL timeout, blob not found |
320
321For full error reference tables, log locations, Log Analytics queries, endpoint metrics monitoring, workspace diagnostics, and the secure workspace setup checklist, see **[references/troubleshooting.md](references/troubleshooting.md)**.
322
323---
324
325## Additional Resources
326
327Detailed reference files for each topic area:
328
329- **[references/networking.md](references/networking.md)** -- VNet, private endpoints, DNS zones, NSG rules, service tags
330- **[references/compute.md](references/compute.md)** -- GPU SKUs, compute instances, clusters, serverless, Kubernetes
331- **[references/endpoints.md](references/endpoints.md)** -- Managed online, batch, Kubernetes, and serverless endpoints
332- **[references/identities-acr-storage.md](references/identities-acr-storage.md)** -- Managed identities, ACR integration, storage accounts
333- **[references/cli-powershell.md](references/cli-powershell.md)** -- Complete az ml CLI and PowerShell command reference
334- **[references/terraform.md](references/terraform.md)** -- Full production-ready Terraform configuration
335- **[references/troubleshooting.md](references/troubleshooting.md)** -- Log reading, debugging, error tables, setup checklist
336
337### External Documentation
338
339- [Azure ML Documentation](https://learn.microsoft.com/azure/machine-learning/)
340- [Azure AI Foundry Documentation](https://learn.microsoft.com/azure/ai-foundry/)
341- [az ml CLI Reference](https://learn.microsoft.com/cli/azure/ml)
342- [Az.MachineLearningServices PowerShell Module](https://learn.microsoft.com/powershell/module/az.machinelearningservices/)
343- [Terraform AzureRM ML Workspace](https://registry.terraform.io/providers/hashicorp/azurerm/latest/docs/resources/machine_learning_workspace)
344- [Secure Azure ML Workspace with VNet](https://learn.microsoft.com/azure/machine-learning/how-to-secure-workspace-vnet)
345- [Managed Network Isolation](https://learn.microsoft.com/azure/machine-learning/how-to-managed-network)
346- [Troubleshoot Online Endpoints](https://learn.microsoft.com/azure/machine-learning/how-to-troubleshoot-online-endpoints)
347- [Azure ML RBAC Roles](https://learn.microsoft.com/azure/machine-learning/how-to-assign-roles)
348- [GPU VM Sizes](https://learn.microsoft.com/azure/virtual-machines/sizes/gpu-accelerated/nd-family)