Self-Driving Data Centers for the Next Decade
Where will they be taking us?
As is my usual practice for the annual predictions column, I queried the official Hot Aisle Insight virtual crystal ball for guidance on trends and developments that dominate in the coming year, especially since 2020 begins a new decade. This time, it did not respond, which concerned me, since I just changed its security settings, allowing it to auto-update its own highly advanced, Hot Aisle Insight artificial intelligence engine. Instead, the orb turned blue, so perhaps it may have inadvertently downloaded the last Microsoft Windows 7 patches. Evidently, it will need more development (by humans) and should not be left unsupervised. Therefore, I was forced to write this article without AI augmentation, using traditional “OK Boomer” research methodology (assuming manually typing in the Google search bar is considered traditional).
While I usually try to predict which technology developments and trends will influence the course of data centers for the coming year, I believe we all need to look at this in a broader perspective — beyond the nuts and bolts of the data center. Clearly, this past year has seen much more general public awareness, avocation, opinions, and polarization regarding the impact and sources of climate change on many fronts, renewable energy, material sustainability, and even types of foods and the food chain. Yet, despite that, society demands ever more technology-driven services, which require tons more IT hardware and, of course, data centers to house all the IT equipment.
New data centers have vastly improved their energy efficiency over the past few years in general, and the “at scale” sites continue to race toward a power usage effectiveness (PUE) of 1.0x. Nonetheless, it is the IT equipment that still consumes massive amounts of power. In partnership with energy producers, largescale data centers are driving and expanding the production and use of renewable energy, such as solar and wind, by way of power purchase agreements (PPA). As a result, data centers are not completely out the crosshairs of climate change and sustainability focused organizations, but things have improved. In 2017, Greenpeace gave favorable ratings to some data center operators for the use of renewable energy in a report titled Clicking Clean (www.clickclean.org).
They are becoming less of a target and are one of the climate change leaders, using technology to optimize both the facility energy efficiency as well as the computing hardware. Google is already utilizing machine learning to directly control and optimize its cooling systems.
The self-driving data center is really the ultimate embodiment of the “lights out” concept for IT management, which has been heralded for over two decades and is now getting closer to fruition. While humans are still handling the physical task, it is now technically possible for robots to replace failed plug-in modules in bladeservers.
Moreover, I continue to be amazed at the ongoing and expanding data center builds, mergers, acquisitions, and divestures across the spectrum of players. This new paradigm encompasses wholesale and retail colocation providers , telecom carriers, and hyperscale cloud services. None of them can seem to build, buy, or lease data centers fast enough to meet demand (or predicted demand), and they all are driven to “the edge” (wherever it is). And, speaking of the edge, we all know that no discussion about the edge is complete without invocating 5G.
The hype of 5G was boundless in 2019 and will only expand in 2020. Consumers can livestream everything in 4K HD from their phones or download movies in seconds instead of minutes, and gamers can walk around with immersive virtual reality (VR) glasses without any lag (and, hopefully, without getting run over by self-driving vehicles). Lest we not forget the infinite world of the IoT and all its iterations from the wide and growing range of voice-enabled devices as well other wireless, smart-devices, such as wearables for fitness and biomedicals, all gathering and sending massive amounts of data to be processed by the AI and ML grist mills in hyperscale data centers (http://bit.ly/3335lez).
These AI processors are struggling to keep up with the workloads and continue to require more power. For example, Google’s third-generation Tensor Processing Unit (TPU) had to be liquid-cooled due to its higher power draw (http://bit.ly/2CZAzbS).
Then there are the endless promises (or threats) of self-driving, fully autonomous vehicles (known as Level 5), which will be enabled by the low latency of 5G networks. The production vehicles currently available (Levels 1, 2, and 3) have features like adaptive cruise control, automatic braking, and lane keeping and warning systems, which are only based on onboard systems and do not communicate with other vehicles, traffic control systems, or edge computing or centralized data centers, not to mention flying taxis, package delivery drones, and other not-so-distant aspects of the daily life of the Jetsons.
Moreover, there are symbiotic ecosystems that have recently expanded, as state and local governments try to bid for massive data center investments after seeing the success of the ongoing development of Data Center Alley in Ashburn and other areas of Virginia. In many cases, the colocation providers negotiated or have been offered tax advantages for the site, which would help lower the final cost to the end-user organization. Significant financial investments and/or job commitments allow wholesale colocation providers to either obtain new tax incentives or qualify for existing programs. Currently, more and more states are offering some form of incentives to bring largescale data center projects into areas that need economic development. At this point in time, more than half of the states in the U.S. have varying forms of tax incentives.
The Bottom Line
This endless cycle needs to find its equilibrium point. Yes, there are numerous benefits that ongoing development and advances in computing and communications technology have brought to daily life as well as education, health, and science. But there is a price to pay, and it will require the consumers of the technology to recognize that their demands for convenience and comforts are part of the problem — not the data center.
We have now fully immersed ourselves in the digital and social lifestyle and remain connected 24/7 through both wired and wireless devices. In fact, use of smartphones and tablets now surpasses the PC for activities like research, communication, and purchasing.
While most of this article focused on big picture issues that are above my pay grade, I cannot help but mention data center cooling, given the name of this column. We have seen a huge shift in the standards, trends, and practices since 2010. The original ASHRAE “Thermal Guidelines” focused on rigidly enforcing 68°F and 50% rh environmental conditions (resulting in PUE of 2.0 or higher). But, now, there’s a growing acceptance to keep data center temperatures at 75° or more with a broader rh range of 40% to 60 % (or implementing dew point control) for enterprise and colo facilities, which results in a massive improvement in energy efficiency without impacting IT reliability.
The hyperscalers have always gone much farther, since they are only responsible to their own operating risks and set their own environmental standards. They are the groundbreakers constantly exploring how far they can go to save cooling energy while meeting ever-increasing power density challenges. The mainstream data center designers, owners, and users eventually see how well these leading edge-cooling schemes (air- or liquid-cooled) work out before implementing them in their own facilities. And, finally, regardless of how close facilities get to a PUE of 1.0, each megawatt (and now gigawatt) is converted to waste heat and dumped into the air, a lake, a river, or even the ocean. More efforts are needed to focus on the reuse of waste heat energy. My prophecy for 2020 through 2030 is perhaps more hope than prediction … unless Mr. Fusion will become commonly available (in different sizes and colors) or the net-zero data center becomes feasible.
With the convenience and benefits that technology brings, it is important to remember that everything has a price — ultimately there is no such thing as a carbon-free lunch.
So, will there be self-driving (or flying) micro data centers to satisfy the demands of autonomous vehicles when traffic jams occur if the onboard processors get an auto-upgrade? Perhaps not, but we will hopefully still be able to access an EPO somewhere (just in case they were right about Skynet).
As this is the time of the year when shopping is a make-or-break event for many businesses, start shopping and stay tuned to see how these predictions pan out. Have a happy, semi-secure shopping experience, be it in-store, on your favorite mobile device, or by using your favorite voice-enable assistant (Alexa, Cortana, Google Assistant, or Siri). Have a green holiday season and a happy, sustainable New Year!