What does the future of the data center look like?
If your answer is “dead”, you would be wrong.
The data center you know today is evolving rapidly. It is fast-forwarding towards becoming a fully autonomous organism, one that is maintained by robotics and drone technology controlled by AI. A confluence of trends is driving this.
For one, demand for IT infrastructure has never been higher. The role out of large distributed applications and platforms is driving demand for capital efficient data centers. These facilities are much larger and built for exceptional reliability and efficiency. Which, in turn, is driving demand for automation and skilled labour.
Outside of large data centers, edge and micro data centers are getting deployed at a breakneck pace to support IoT, sensor telemetry aggregation, video surveillance, factory automation, military applications, and more. These data centers are typically small, self-contained, and in locations that are not easily accessible. There is no better poster child for this than the underwater data center pods Microsoft rolled out recently.
Microsoft’s Northern Isles underwater data center, which was retrieved from the seafloor off the Orkney Islands in Scotland after two years. Photo by Simon Douglas.
Another trend is the disaggregation of data center compute and network hardware. Disaggregating allows for the creation of pools of capabilities to deliver the needed compute, network, and storage for an application. Technologies like PCIe over Fibre and advancements in on-chip optical interconnects make this possible. We also see examples of this in Reference Architectures like Rack Scale Architecture from Intel. Or projects like Shoal from Microsoft, which is focused on interconnect fabric design. The tip of the spear are cloud providers like Microsoft Azure, and data center providers like Vantage Data Centers who operate large-scale infrastructure for hyperscalers. The continued adoption of disaggregation will drive manufacturers towards hardware designs that are not maintained by people, but by robotics.
We are seeing a similar approach in application design and deployment patterns. New applications are built to be decoupled from the underlying hardware and operating systems. This frees operators to update hardware and infrastructure without taking critical applications down and, in most cases, removes the need for immediate hardware attention. Platforms like Function-as-a-Service (FaaS) and Kubernetes are driving and accelerating this trend.
Unprecedented access to and affordability of sensors, robotics, drone technology, and software is driving adoption by organizations across industries to improve their business outcome. We see examples of this all over, from factory and warehouse automation, to drone delivery. The same is happening within data centers to realize greater performance, lower costs, and less risk.
Lastly, human error continues to be the leading cause for downtime as found in research by the Uptime Institute, Gartner and others. A key advantage of an autonomous data center system is that it can codify best practices, continuously learn, and optimize from experience. This significantly reduces the number of human touchpoints, along with the risk of downtime. It also frees up an organization’s professional resources to focus their efforts on more high-value work.
Although these trends are shaping the future of data centers, we are still in the early stages. Yes, the hyperscalers have experimented and adopted many of these trends, but the rest of the industry is lagging behind. As the benefits become clearer, and current investments are depreciated, we can expect adoption of the next-generation data center to accelerate.