Microsoft continues to expand Azure Arc’s capabilities to transform it into a hybrid cloud and multi-cloud platform. At the recent Spring Ignite conference, Microsoft announced the general availability of Azure Arc enabled Kubernetes, and the preview of Arc enabled machine learning.
Azure Arc - The Cornerstone of Microsoft Hybrid and Multi-Cloud Strategy
Initially announced in 2019, Azure Arc is a strategic technology for Microsoft to expand its footprint to the enterprise data center and other public cloud platforms. Azure Arc is the only offering available in the market to manage both the legacy infrastructure based on physical servers and modern infrastructure powered by containers and Kubernetes.
With Azure Arc enabled servers, customers can onboard existing Linux and Windows servers running on bare metal servers or virtual machines to Azure Arc to manage them centrally. These servers could be running in on-premises environments or public cloud environments. Once registered with Azure Arc, they can seamlessly extend the Azure-based automation, management, and policy-driven configuration to any server irrespective of their deployment environment. This simplifies the fleet management and governance of infrastructure.
For example, with Azure Arc enabled servers, DevOps teams can roll out a consistent password policy to all the machines running in Azure VMs, on-prem data center, and even to Amazon EC2 or Google Compute Engine instances. They can also audit the compliance and remediate the issues from a centralized control plane.
Azure Arc enabled Kubernetes lets customers register Kubernetes clusters with Azure to take control of the cluster sprawl. Similar to Azure Arc enabled servers, they can apply consistent policies across all the registered clusters. An additional advantage of Azure Arc enabled Kubernetes is the integration of the GitOps-based deployment mechanism. Cluster managers can ensure that every Kubernetes cluster runs the same configuration and workloads across all registered clusters. GitOps provides at-scale deployment of workloads spanning the clusters running in the public cloud, data centers, and the edge.
Azure Stack, the hardware-based hybrid cloud offering from Microsoft, runs both VMs and managed Kubernetes clusters that can be registered with Azure Arc.
Optionally, Azure Arc customers can ingest the logs and metrics from servers and Kubernetes clusters into Azure Monitor - an integrated observability platform.
As of March 2021, Arc enabled servers and Arc enabled Kubernetes offerings are generally available
Exploiting Kubernetes to Bring Managed Services to Azure Arc
Kubernetes has become the level playing field for running modern workloads. It’s transforming to become the new operating system for running distributed workloads, including databases and machine learning platforms.
Kubernetes plays a crucial role in Azure Arc by becoming the infrastructure foundation for running managed services such as databases and machine learning. Microsoft is leveraging Kubernetes to abstract the low-level infrastructure to run platform services reliably. Azure Arc enabled data services and Azure Arc enabled machine learning are early indicators of how Microsoft plans to unleash its managed services to run on any Kubernetes cluster.
Azure Arc enabled data services extends Microsoft Azure’s managed databases, including PostgreSQL Hyperscale and SQL Managed Instance to Kubernetes clusters running in hybrid and multi-cloud environments. Customers can use Azure Portal or the CLI to manage the lifecycle of database servers deployed through Arc enabled data services. The key advantage of this service is the ability to run databases in disconnected environments such as edge locations. Customers can run the databases in a highly secure environment without opening any outbound connections to the cloud.
Having experimented with databases, Microsoft is all set to bring machine learning to Azure Arc. Customers get the familiar Azure ML experience running in on-prem environments and other public cloud environments. Arc enabled machine learning combines the best of Kubernetes with data science and machine learning workflows. DevOps teams can provision workspaces with pre-configured Conda and Jupyter Notebook IDE. Through Role-Based Access Control (RBAC), data scientists and ML engineers can be given access to select operations needed for their job. With Arc enabled machine learning, customers can mix and match CPU hosts and GPU hosts of a Kubernetes cluster to run distributed training jobs. The models can then be deployed in managed Kubernetes clusters in the cloud or at the edge for inference.
Arc enabled machine learning is a masterstroke from Microsoft. It essentially brings ML Platform as a Service (PaaS) closer to the origin of the data. Customers may have large datasets uploaded to Amazon S3 while the ML training jobs are running in Azure. In that case, they can launch an Amazon EKS cluster in AWS to run Arc enabled machine learning with the same Jupyter Notebook and Azure ML SDK to train a model on AWS. The machine learning model can then be registered and deployed in Azure ML for inference.
Microsoft’s investments in Azure Stack-based hardware and Azure Arc platform become the critical differentiating factor. Azure is the only public cloud platform with hardware and software-based choices for implementing an enterprise hybrid cloud and multi-cloud strategy.