Answering the question “What is multicloud” seems simple enough: It’s any cloud environment that involves more than one cloud, right? Well, maybe. If you dig deeper into the definition of multicloud, things can get murkier. There are different ways of defining what counts as a distinct cloud. The lines between hybrid cloud and multicloud can be blurry, especially in today’s complex hybrid cloud landscape. And there is some debate about the extent to which multiple clouds need to integrate with each other to qualify as multicloud.
With these nuances in mind, here’s an overview of the different ways of answering the “What is multicloud” question.
A Basic Definition of Multicloud
Again, at a high level, a multicloud architecture is defined as any type of cloud architecture that includes more than one cloud. The classic example of multicloud is an architecture that includes services from more than one public cloud, like AWS and Azure, but other approaches are possible.
Multicloud architectures have existed in practice among some organizations for many years. As early as 2010, it was possible to find companies using multiple public clouds at once, or using a hybrid architecture that (depending on how you define multicloud) fit within the multicloud mould.
Circa 2017, however, multicloud turned into something of a buzzword, as more businesses turned to multicloud as a way to reap benefits like lower costs and greater reliability.
Challenges in Answering the “What Is Multicloud” Question
The exact nature of the multicloud architectures that businesses have built during the past several years, however, come in many forms. The definition of what counts as multicloud and what doesn’t hinges on how you choose to think about the key variables within the various possible configurations that leverage multiple clouds in one way or another.
One major variable in the “What is multicloud” question is the extent to which workloads that are spread across multiple clouds are integrated with each other.
You could consider multicloud to include a strategy where a business runs one application in one cloud (like AWS) and runs an entirely separate application on another (like Azure). In this case, the business would certainly be using multiple clouds.
But, because those clouds would not be integrated with each other in any specific way, you could argue that they don’t really constitute a multicloud architecture. They are just two completely separate clouds that happen to be used by the same business.
If you take the latter view, you might define multicloud in a way that requires some level of integration between workloads. You might argue that an application that is hosted on one cloud, but ingests data that is hosted on another cloud, counts as a multicloud architecture because in this case the multiple clouds are actually connected in a direct way.
Another factor in defining multicloud is whether workloads are tied to a specific cloud platform, regardless of how many clouds are actually hosting them at any given time. You could contend that in order to count as multicloud, your architecture needs to free you from complete dependence on any single cloud.
For example, a virtual machine image that is designed in such a way that it can easily be ported among AWS, Azure, GCP or any other major cloud would qualify as an application that leverages a multicloud architecture--even if it’s only hosted on one cloud at a time. An approach like this achieves the flexibility that multicloud architectures are intended to provide. It involves the potential use of multiple clouds, albeit not necessarily at the same time.
Multicloud and private clouds
Deciding whether and how private clouds could fit within a multicloud architecture is also a thorny issue.
If, for instance, you build an OpenStack private cloud and deploy it on AWS public cloud infrastructure, you could view that as a multicloud setup because it includes multiple clouds: a private one and a public one.
Or, you could argue that it’s really just a private cloud that happens to be hosted in a public cloud data center. It doesn’t involve spreading workloads across a private and public cloud as much as layering one cloud on top of the other.
You could alternatively view a private-on-public setup as a form of hybrid cloud, which brings us to the next variable.
Hybrid cloud vs. multicloud
Perhaps the most contentious issue of all when answering the question of “What is multicloud?” is where hybrid cloud fits in. Most definitions of multicloud architectures (like this one and this one, for example) treat hybrid cloud and multicloud as distinct types of architectures. They suggest that multicloud requires the use of multiple public clouds, while hybrid cloud involves using a public cloud and either a private cloud or simply on-premises servers that are not configured to run as a cloud at all.
One could argue, however--as some folks do--that hybrid cloud is one possible form of multicloud. If you define multicloud as any environment involving at least two clouds, then a hybrid environment that mixes private cloud resources with a public cloud would qualify as multicloud.
Spelling out the difference between hybrid cloud and multicloud gets even more complicated when you consider that there are a variety of approaches to building hybrid clouds.
You could deploy on-premises servers and integrate them with a public cloud service in some way to build a hybrid cloud. Or, you could use a framework like AWS Outposts or Azure Arc to host public cloud services on your own infrastructure. These are fundamentally different approaches to hybrid cloud: The first involves two distinct environments that simply interface with each other, while the second seamlessly extends public cloud services across private and public infrastructure.
Maybe the latter setup doesn’t qualify as multicloud at all because it involves just one cloud--the public cloud that powers the hybrid architecture. But you could also view this scenario as fitting the multicloud definition because it involves two distinct clouds--your on-premises environment and a public cloud.
Or, maybe you believe that a framework like Outposts doesn’t give you a multicloud environment because everything is ultimately tied to a single public cloud platform. However, you would think it would count as multicloud if you used a third-party framework, like AppScale, to run public cloud services on your own infrastructure.
In short, there is an array of ways to build hybrid architectures. Whether any or all of them fit into the multicloud model depends on your point of view.
Conclusion: Is There One Answer to the “What Is Multicloud” Question?
Ultimately, the many nuances at stake in defining multicloud may not matter. What’s really important is not what you choose to call your cloud architecture, but rather achieving whichever goals are most important to you.
If cost optimization is your main priority and a hybrid architecture helps you do that, then you’ve achieved your goal, regardless of whether you consider the hybrid cloud a multicloud. If you set up your applications to fail over automatically from one cloud to another to maximize reliability, that’s what ultimately matters--whether you choose to call this setup multicloud is not really important.
The bottom line: There are many answers to the “What is multicloud” question, and all of them are valid. Don’t worry too much about whether your architecture meets others’ definition of multicloud--as long as it meets your organization’s technical goals.