At least in the U.S., the discussion about clouds is decidedly U.S.-centric. Pundits talk about the cloud choices as if the only clouds to consider are AWS, Azure, Google Cloud, and Oracle. This creates unintended boundaries on intelligent cloud dialogue, as many people tend to think about a single cloud that can service all needs.
But while this might be true for companies whose businesses exist within a fairly narrow geographical area, any company with a multinational presence will find that limiting the cloud discussion lacks the nuances necessary to appropriately plan a meaningful path forward.
The Role of Proximity
The years-old definitions of cloud as a set of centralized resources for handling workloads have largely been debunked. More modern cloud discussions no longer revolve around where the resources reside, choosing instead to focus on the fungibility of resources in support of workloads that might run anywhere. In fact, this is a key tenet behind the notion of availability zones, allowing enterprises to split capacity across different pools in the event that one of those pools is not available (Amazon’s S3 outage in 2017 is a good example of why this matters).
But the treatment of cloud as something other than a centralized set of resources is important for reasons other than availability. For applications that have even a small degree of latency sensitivity, proximity matters. If a company’s reach is contained within narrow boundaries, this might not matter. But for multinationals with a global presence, there will clearly be a need for clouds that are closer to the user.
And, of course, as the notion of user expands to include more than just people but also connected devices, the proximity requirement becomes even more important. Imagine remote mining facilities or monitoring and control systems in hard-to-reach environments.
Whatever the reason, the conclusion is that there will be a market for distributed clouds. And that market will bring location into play, granting advantage to those with a local presence. While it could be that every cloud provider expands its cloud presence into every major geography, a more likely outcome is that regional cloud offerings emerge. And if reach becomes a differentiator, it seems unlikely that all areas will be served by the same cloud.
In this scenario, the need for workloads near users will drive companies to pursue multiple cloud providers. Still unsure? Consider connectivity services purchased from the large telco providers. How many large companies have a uniform offering across a diverse geographical operating environment? And that means the future of cloud is unlikely to stop at cloud — no, multicloud seems inevitable.
Data privacy is increasingly critical, particularly as breaches become both more frequent and more devastating. The European Union’s General Data Protection Regulation (GDPR) is certainly a reflection of what will become more tightly monitored if not regulated space.
But even with GDPR, there will be industry-specific regulations that mandate protections, especially around data that is of security interest. In some cases, enterprises will find that tight control over data, being mindful of national borders, provides the right balance of risk mitigation and control.
Here again, visions of a centralized cloud are likely to be supplanted by a distributed set of resources provided by a more diverse set of suppliers than most single-cloud discussions account for.
All the Other Reasons
And then there are the more mainstream reasons that people might consider a multi-solution approach to their cloud ambitions. Cloud providers will ultimately be treated like other technology suppliers, which means procurement teams will favor multivendor strategies to create economic leverage as the price for more and more workloads becomes meaningfully large relative to a company’s overall IT spend.
Applications might exist only in one or another cloud. Microsoft and Oracle, for example, seem poised to locate their valuable applications in their own clouds, which means companies whose application needs span more than one provider will be forced to move to multicloud.
The clouds themselves will have their own capabilities. Google has been aggressively pursuing its machine learning use cases with the introduction of Tensor Processing Units (TPUs). Amazon has its own ambitions around efforts like Lambda and Greengrass. Microsoft is pursuing edge possibilities with AzureStack. These will all be useful for different things, which will create still more pressure for companies to adopt a multicloud approach to IT.
Multicloud but Single Operations
If the future is likely multicloud, the question for enterprises will be whether that necessarily leads to a fractured operations environment. Because multicloud is still early in its adoption, the tools required to span operations (monitoring, provisioning, remediation, and so on) are still being developed. But with the need very clear and the implications so important, it seems obvious that the set of multicloud building blocks will expand over time.
The broad support for Kubernetes, for example, bodes well for a future that spans multiple clouds. The key for enterprises will be in making sure that the architectures they settle on for individual cloud efforts do not ultimately preclude them from participating in a more multicloud future. Minimally, enterprises will do well to add multicloud as an explicit consideration to their architectural planning. More ideally, they will begin their efforts in earnest to identify and deploy infrastructure spanning all of compute, storage, networking, and tools, remembering that cross-domain efforts like security and automation often fall in the gaps.
Feature image via Pixabay.