Reports of the imminent death of the data center have been greatly exaggerated. That’s not to say that data centers can remain static, resembling the ones of ten or even five years ago. Thanks to the explosive growth in cloud computing, and the increasing numbers of enterprises adopting a “cloud first” strategy, many organizations are having to fundamentally rethink how they architect and use their on-premises data centers. That’s paving the way for a transformation in the infrastructure within them, and also a transformation in the way that they are managed.
Cloud Spending Explosion
To get an idea of what data centers are up against, Gartner forecasts that spending on public cloud services worldwide will grow 18.4% in 2021 to total $304.9 billion, up from $257.5 billion in 2020. As a proportion of total IT spending, the research house expects cloud to make up 14.2% of total global enterprise IT spending in 2024, up from 9.1% in 2020.
But these figures mask much greater migrations to the cloud. Gartner also says that many companies are allocating 40% or more of their IT budgets to cloud and cloud-related services. It’s not uncommon to find large enterprises — General Electric is one example — moving 60% of their workloads out of the data center and into the public cloud. And, there are examples of enterprises that have moved a higher proportion still.
Yet corporate data centers still remain, for a variety of reasons. The most obvious of these are that some applications are not suitable for migration to the cloud for regulatory, latency, or security reasons. This last reason is rapidly becoming irrelevant as cloud operations frequently offer better security than is possible in a corporate data center.
“The role of the traditional data center will be relegated to that of a legacy holding area, dedicated to very specific services that cannot be supported elsewhere, or supporting those systems that are most economically efficient on-premises,” says David Cappuccio, a Gartner analyst.
Data Center Inefficiencies
One very basic result of migration to the cloud is that many existing corporate data centers now require fewer compute resources running in them. That means that they may well be overbuilt in terms of space, HVAC capabilities, and even power supplies. That’s likely to make them less efficient, when compared to the sorts of environments that cloud operators like AWS keep their infrastructure in.
But cloud operations have also highlighted how inefficient the traditional application-driven data center design really is. This data center model is essentially one of multiple computing infrastructures, each dedicated to a different application, that all happen to share the same physical location.
It’s also inefficient in terms of individual pieces of equipment. That’s because the traditional data center uses a multiplicity of hardware, with the most suitable or “best” servers and storage systems being chosen for each given application. Equipment refresh rates may be in the region of three to seven years, and maintenance — simply keeping hardware running — is a non-trivial exercise in terms of time and of cost.
Frequent Refresh Cycles
Contrast this with the cloud, where homogenous equipment environments and high levels of automation mean operational efficiency is far higher. This allows cloud operators to maintain aggressive two- or three-year equipment refresh cycles, taking advantage of increased scales and performance offered by newer generations of technologies and spending far less time and money on maintenance and break-fixes of aging equipment.
Data Center Evolution
So how will data centers evolve to become an efficient “legacy holding area,” as Gartner’s Cappuccio puts it?
The first two things that are likely to change are the size and location of modern data centers. Clearly as workloads move to the cloud there is less need for space, so data centers will shrink. But, more importantly, many will be moved closer to where they need to be as the trend towards edge data centers gathers steam. That means more, smaller, data centers closer to factories, branch offices, or customers. There’s an IT saying that “slow is the new down”, and it’s because of the need to avoid “slow” in IoT applications, customer-facing applications, and many other areas that these smaller edge data centers will pop up in places far away from the nearest AWS or Azure cloud data center.
Mimicking the Cloud
It’s also what’s inside data centers that will also change dramatically. That’s because to remain viable compute environments they will have to learn from, and, to an extent, mimic, the way that efficient cloud environments are set up.
What that entails is a huge increase in the use of commodity hardware to simplify and streamline hardware operation and management, and a converged infrastructure with pools of virtualized servers, storage and networking capacity, and maybe pools of container hosts, all managed and orchestrated as far as possible from a single pane of glass.
If that sounds like data centers will look more like public clouds, it’s because that’s exactly what is happening. Expect to see more enterprises with private and hybrid clouds; more, smaller, geographically dispersed data centers; and far fewer of the old-fashioned giant, siloed, and above all, inefficient data centers of the past.
Read next: Best Server Virtualization Software of 2021