Hybrid cloud tales from the field: How is Azure Arc being used?

Cloud migration, enterprise systems consolidation, and agility are among the big reasons that early adopters are looking at hybrid cloud. But “doing hybrid right” requires different approaches to architecture and development because you are working with a mirror image of the cloud.
Written by Tony Baer (dbInsight), Contributor

Head to the cloud or bust. That's one conclusion that could be drawn from continued double digit growth of cloud computing over the past year: AWS revenues jumping 25% while Azure and Google Cloud saw steeper 50% marks. With the pandemic driving enterprises to pivot their businesses to the new facts on the ground, it's not surprising that the long term secular trend of cloud adoption an added shot in the arm.

But for many organizations, public cloud deployment is not practical. Constraints could include data residency laws, lack of public cloud presence in certain geographies, and/or highly stringent near real-time or hard real-time use cases where network roundtrips to the public cloud would make latencies intolerable.

Yet, these organizations are still looking for gaining the advantages of operational simplicity, flexibility, and agility for those very use cases that cannot easily move into the public cloud. That has sparked emerging demand for hybrid cloud alternatives – but as we noted in our research last year, there are many forms of hybrid cloud, from software-defined hyperconverged environments to full-blown turnkey systems, and from customer- to vendor-managed. Hybrid cloud platforms may be out on the market, but customers are still on a learning curve. It's still early days for adoption.

So, when Microsoft reached out to make several early customers of Azure Arc available to share their experiences, we jumped at the chance. We listed a couple them in our close-up look at Arc last fall. They included Siemens Healthineers, a medical diagnostics systems provider and Africa's Talking, a provider of mobile payments solutions with an accompanying development environment. Rounding out the list is Ferguson, a residential plumbing, HVAC, and appliances provider that in recent years has expanded beyond its roots from wholesale distribution to a business that also has a national retail network.

Each of these customers are still very early in their rollout of Arc; for instance, Africa's Talking only has a single instance that is live, but is ultimately interested in growing it to 15 - 20 points of presence in countries that it serves.

To recap, Azure Arc is Microsoft's software-defined implementation of the control plane used in the Azure public cloud. It's available in a variety of configurations, some of which require Kubernetes (K8s) expertise, and some of them don't. That could be an important consideration for enterprises looking at adopting hybrid or private cloud platforms because, arguably, K8s skills are more prevalent in the tech vendor community than across most enterprise IT departments. Speaking with early adopters, it's notable that not all of them are yet working with K8s.


Medical imaging systems such as CT and MRI scanners typically carry significant compute power. Siemens Healthineers characterizes them as "edge" devices, although these are in an entirely different class compared to the smart sensors that are typically thought of as edge in IoT networks. When they developed a new digital healthcare application to make these systems smarter, a cloud-based service was the logical path for making it economical to deliver, maintain, update, and control such a service extending to thousands of these edge systems.

But public cloud deployment would not make sense. Most customers – hospitals and clinics – consider this data too sensitive to process or store in a public cloud, not to mention that the latencies of round-tripping such data to and from the cloud would have proven unacceptable. A hybrid cloud solution proved the logical choice, and since Siemens Healthineers was already running SaaS services in the Azure public cloud, Azure Arc provided the means to extend that same control plane to thousands of edge imaging systems. It implemented Arc by using a "receiver" at each customer site to act as the gateway between the diagnostic system(s) and the Azure cloud.

Like Siemens Healthineers, Africa's Talking is also using Arc to deliver an as-a-service offering to its customers, but that's where the similarity ends. The company developed a mobile payments PaaS platform as an underlying engine for developers to localize and support their own apps, such as video chat and other real-time services, for national markets. The purpose of Arc is making it possible to implement and streamline maintenance of cookie cutter local points of presence, replacing the custom scripts that the company currently utilizes. And with standardized maintenance, the company could redirect resources toward expanding its portfolio of services. The local points of presence are out of necessity; the Azure public cloud has only one region on the continent (in South Africa), and some countries, such as Rwanda and Egypt, have strict data sovereignty laws.

At Ferguson, Arc is both a steppingstone to public cloud and a means for standardizing the retail customer experience across its brick and mortar locations. For now, the company uses Azure as a common control plane, connecting the on-premises environment to the same management console that they also use for applications and databases already running in the Azure public cloud. Aside from Azure Portal, they are seeking to leverage other Azure operational tools such as Azure Site Recovery for disaster recovery and Azure Migrate for lifting and shifting applications the Azure Public Cloud. Ferguson has utilized both Azure Arc for Data and Azure Arc-enabled servers for simplifying monitoring and management of their hybrid environment. This is accomplished through Azure Arc's agents, which connect these on-premises systems to the Azure control plane.


Just as the use cases are varied, so are the hurdles. For instance, both Siemens Healthineers and Africa's Talking were already working in the cloud – it was not a matter of adapting on-premises workloads. Both faced the challenges of deploying to thousands of heterogeneous targets. It's the exact opposite of deploying to a public cloud where the infrastructure (and ownership) is more standardized. There's an irony here – the advantage of cloud deployment is operational simplification, yet when it comes to hybrid cloud deployments, the challenge is ensuring that you don't simply paper over the differences. Specifically, automating deployment and operations to realize economies of scale over thousands of endpoints at customer sites requires special attention to design and development processes.

At first glance, this might appear to be more of a concern for a SaaS vendor seeking to telescope deployment across multiple customers. But the challenges could also be similar for enterprises with highly distributed geographic computing footprints.

For Siemens Healthineers, it dictated major changes to the routine process of pushing new functionality and updates to a single standardized target, but to thousands of endpoints. That meant changing from a standard "push" of new or updated functionality to a public cloud to, what the company terms a "GitOps" approach where deployments; for instance, specifications are pushed into the cloud, with the localized Arc servers then pulling down those changes either at specific points in time or only after the customer approves.

For Africa's Talking, it's the challenge of harmonizing implementations in a highly heterogeneous business and technology environment. Its local points of presence are collocated inside the data centers maintained by mobile telcos in each of the nearly 20 countries that the company serves. The infrastructure will vary and so will ownership; in some cases, Africa's Talking maintains the local Azure cloud environment in the data center, while in others, the telco runs it and Africa's Talking operates as a guest. Among the steepest challenges are building unified views of each of the K8s instances, each of which have different APIs, across network connections that are not always reliable.

As a diversified company, Ferguson's challenge was not specific to hybrid cloud per se, but getting a common view of data and applications across 5000 virtual machines across two datacenters. The challenge is where to start. Ferguson initially focused on data to create a centralized repository for many aspects of its businesses. Additionally, they found the ability to expand their Azure footprint by taking advantage of services such as Windows Virtual Desktop, Azure NetApp Files, and Azure Arc. With Arc, they could use Azure Portal as a common control console from which they could manage these services.

A key lesson learned there has been the urgency for governance. Governance is challenging enough when you run systems on bare metal, but when you introduce virtualization, that brings new complexities. Unless you're running on bare metal, virtualization is how the cloud can efficiently apportion compute. It means having to track what's running where understanding whether SLAs are being met and whether data is being utilized properly, and only by those authorized to access it.

Governance for hybrid cloud is even more complicated when your customers – whether they are independent companies or separate business units – own the VMs, containers, or microservices. Governance grows essential for enforcing rules at scale and ensuring that problems are properly escalated. According to Thomas Friese, who is responsible for the digital platform at Siemens Healthineers, lacking governance, the consequences would be "overwhelming" for the operations team.

Disclosure: Microsoft is a dbInsight client.

Editorial standards