Add usage pattern for on-prem with cloud extension

Author: fbeltrao

articles/machine-learning/how-to-attach-kubernetes-anywhere.md

@@ -57,6 +57,7 @@ With Arc Kubernetes cluster, you can build, train, and deploy models in any infr
 | Usage pattern | Location of data | Motivation | Infra setup & Azure Machine Learning implementation |
 | ----- | ----- | ----- | ----- |
 Train model in cloud, deploy model on-premises | Cloud | Make use of cloud compute. Either because of elastic compute needs or special hardware such as a GPU.<br/>Model must be deployed on-premises because of security, compliance, or latency requirements | 1. Azure managed compute in cloud.<br/>2. Customer managed Kubernetes on-premises.<br/>3. Fully automated MLOps in hybrid mode, including training and model deployment steps transitioning seamlessly from cloud to on-premises and vice versa.<br/>4. Repeatable, with all assets tracked properly. Model retrained when necessary, and model deployment updated automatically after retraining. |
+| Train model on-premises and cloud, deploy to both cloud and on-premises | Cloud | Organizations wanting to combine on-premises investments with cloud scalability. Bring cloud and on-premises compute under single pane of glass. Single source of truth for data is located in cloud, can be replicated to on-prem (i.e., lazily on usage or proactively). Cloud compute primary usage is when on-prem resources aren't available (in use, maintenance) or don't have specific hardware requirements (GPU). | 1. Azure managed compute in cloud.<br />2. Customer managed Kubernetes on-premises.<br />3. Fully automated MLOps in hybrid mode, including training and model deployment steps transitioning seamlessly from cloud to on-premises and vice versa.<br />4. Repeatable, with all assets tracked properly. Model retrained when necessary, and model deployment updated automatically after retraining.|
 | Train model on-premises, deploy model in cloud | On-premises | Data must remain on-premises due to data-residency requirements.<br/>Deploy model in the cloud for global service access or for compute elasticity for scale and throughput. | 1. Azure managed compute in cloud.<br/>2. Customer managed Kubernetes on-premises.<br/>3. Fully automated MLOps in hybrid mode, including training and model deployment steps transitioning seamlessly from cloud to on-premises and vice versa.<br/>4. Repeatable, with all assets tracked properly. Model retrained when necessary, and model deployment updated automatically after retraining. |
 | Bring your own AKS in Azure | Cloud | More security and controls.<br/>All private IP machine learning to prevent data exfiltration. | 1. AKS cluster behind an Azure VNet.<br/>2. Create private endpoints in the same VNet for Azure Machine Learning workspace and its associated resources.<br/>3. Fully automated MLOps. |
 | Full ML lifecycle on-premises | On-premises | Secure sensitive data or proprietary IP, such as ML models and code/scripts. | 1. Outbound proxy server connection on-premises.<br/>2. Azure ExpressRoute and Azure Arc private link to Azure resources.<br/>3. Customer managed Kubernetes on-premises.<br/>4. Fully automated MLOps. |