We are living in an era of cloud modernization, where enterprises are becoming far more nuanced and discerning about how they use public cloud services. This is not to suggest that we’re experiencing full-scale cloud repatriation; a few companies have left the public cloud altogether in favor of conventional on-premises IT, but they’re definitely in the minority.
What’s actually happening across the wider industry is more complex—and far more interesting. We’re seeing a blurring of the lines between what’s on-premises and what’s public cloud. This even includes the hyperscalers expanding their portfolios of on-premises cloud service offerings, as Amazon Web Services recently did with the launch of its new AWS Dedicated Local Zones.
In some ways, it’s a “full circle” moment for the cloud. It’s worthwhile to stop and examine where we came from, how we ended up here today and what may be coming next.
How the cloud became ”the cloud”
Cloud computing started out as a consumption model, not a specific technology or architecture. It all began when companies with infrastructure capacity to spare began selling that capacity to customers as a service. Hence, the Infrastructure as a Service (IaaS) business model was born.
Before IaaS, businesses managed their own IT infrastructure on-premises because they’d never been offered an alternative model. After IaaS, they had the option to pay someone to manage IT infrastructure on their behalf. Many of them were happy to do this because they knew it would free them up to focus on their core business.
Since then, cloud terminology has continued to evolve alongside the changing business needs of customers. For instance, the term public cloud was coined to refer to any services provided in a multitenant environment; over time, it became synonymous with services provided by hyperscalers like AWS, Microsoft Azure and Google Cloud. These public cloud services came to completely dominate enterprise IT, and we had good reason to believe their growth trajectory wouldn’t taper off anytime soon.
However, we’re seeing more and more businesses recognize that the public cloud is not right for all their needs. While these companies generally like the simplicity of the cloud business model, they have valid concerns that going all in on public cloud could lead to higher costs, limited flexibility and regulatory risk.
How we got where we are today: More cloud services in more places
It’s important to note that just because we’re seeing the beginnings of a shift away from public cloud doesn’t necessarily mean we’re going to see a shift back toward on-premises. Many of the issues that led people away from on-premises IT in the first place—such as high CAPEX spend and a lack of scalability—haven’t gone away.
Instead, what’s happening is that many businesses are trying to simultaneously take advantage of the things they like about the cloud and the things they like about on-premises, while minimizing the disadvantages of both. Adopting hybrid cloud is one way they’re doing that, but even the on-premises portion of a hybrid cloud environment doesn’t always look like the traditional on-premises data centers that some of us remember from the days before cloud.
The new approach to on-premises can take several different forms. One example is on-premises services provided by the hyperscalers themselves, as mentioned earlier. It could also be a private cloud, deployed using either on-premises hardware that the business owns and manages itself or managed infrastructure services from a colocation provider.
Today’s enterprises choose many different varieties of on-premises infrastructure because they have so many different challenges they’re attempting to solve for:
One of the early drivers of private cloud adoption was that many businesses wanted a cloud-like experience from their IT—meaning all the scalability, flexibility and reliability that public cloud services offer—but they also wanted increased control over the cybersecurity capabilities they applied to protect their most sensitive workloads.
Others wanted to run cloud services inside a dedicated, single-tenant environment because they were concerned the “noisy neighbor” effect might negatively impact the performance of latency-sensitive workloads such as financial transactions and AI inference. Many businesses that use the public cloud also use the public internet to connect to it, which can further compound their challenges around latency.
Another issue that’s increasingly top of mind for many businesses is regulatory complexity around issues like data sovereignty and privacy. These new regulations have placed limits on what data a company can move, where they can move it and for what purposes. To ensure compliance with these regulations, businesses often have to take their cloud consumption model and move it into an on-premises environment, where they can exercise greater control over their data sets. In fact, AWS named regulatory requirements as a major driver behind its decision to release AWS Dedicated Local Zones.[i]
Finally, many businesses are starting to think about how they can integrate new technologies with high compute requirements into their cloud strategies. A perfect example is AI. Many companies would like to be able to use the public cloud to get the capacity they need for AI training, but they know that doing so would mean giving up control over their proprietary data—something they may not be willing to do.
Instead, these companies are starting to learn that they can take a private approach to AI, and accessing cloud services from inside an on-premises environment could certainly play a role in their private AI strategy.
What’s next: The lines continue to blur
Instead of asking whether the shift back to on-premises in its various forms will continue into the future, perhaps the better question is: Does it even matter?
By this, I mean we already have a number of technologies designed to make private cloud services feel less like traditional on-premises services:
- We’ve got hypervisors that abstract the hardware layer, allowing you to run virtual servers on your own hardware without experiencing the typical limitations.
- We’ve got container orchestration systems like Kubernetes that abstract the OS layer. Everything under Kubernetes is essentially a commodity—it’s treated the same, regardless of whether it’s real hardware or virtual.[ii]
As these technologies continue to become more widely available, I believe the distinction between public and private will become increasingly meaningless.
On top of this, businesses can also institute a cloud adjacent data architecture with the help of a colocation partner such as Equinix. This involves bringing together everything they have inside their on-premises environment—including private cloud services—and pairing it with multicloud access via private, low-latency cloud on-ramps.
Cloud adjacent data enables the future of cloud services—on-premises and otherwise
At Equinix, we believe that a cloud adjacent data architecture is the next logical step in the development of hybrid multicloud, because it’s the answer so many companies have been searching for. It gives them the flexibility to use on-premises without the drawbacks of on-premises and public cloud without the drawbacks of public cloud. It can help address the costs and complexity businesses may have experienced when trying to do hybrid multicloud in the past, because it ensures data can move into and out of a particular public cloud any time the need arises.
As the diagram below shows, the Equinix approach to cloud adjacency is based around the authoritative data core. This is a storage environment that allows you to move your data into different clouds as needed, while still maintaining custody over all your data. You won’t run into the challenge of high data egress fees, as you might if you stored your data directly with a public cloud provider.
The cloud adjacent data architecture takes full advantage of Equinix digital infrastructure services, including Equinix Fabric® and Equinix Network Edge, to move data over private, low-latency connections—either to the public clouds or to the digital edge. In addition, Equinix Metal® offers single-tenant compute and storage capacity via a Bare Metal as a Service (BMaaS) model. Equinix Metal can provide on-demand compute capacity with a cloud-like experience outside the public cloud, as shown in the diagram above. It can also host a variety of on-premises offerings from the hyperscalers in the Equinix partner ecosystem.
To learn more about the Equinix approach to bridging the gap between private and public cloud services, read our guide to cloud adjacent data and storage.
[i] Dan Swinhoe, AWS launches Dedicated Local Zones on-premise cloud offering, Data Center Dynamics, August 24, 2023.