SCRUTINY® silo-based datacenter technology cuts the 10-year total cost of ownership (TCO) of a 16 megawatt (MW) hyperscale datacenter by more than $1 billion, starting with faster time-to-revenue and typical savings of $80 million in construction costs. With SCRUTINY® technology, a complete 16 MW datacenter can cost a fraction of what it costs for a conventional 2 MW data hall, despite having 8 times the capacity.
Moreover, a silo-based datacenter may provide 50% higher performance per watt (a proxy for revenue per watt), compared to a conventional air-cooled hyperscale datacenter having a world-class PUE as low as 1.1 (which most datacenters cannot achieve anyway). Compared to datacenters with a PUE higher than 1.1 (i.e., virtually all datacenters), the performance-per-watt advantage is even greater.
Note that the performance-per-watt advantage of the SCRUTINY® silo-based datacenter technology does not originate with its ultra-low PUE of 1.008 -- its low PUE simply ensures that the available energy goes toward computing rather than being wasted as heat. The major benefits actually stem from advances in power, cooling, and packaging -- all enabled by miniaturization.
Our fundamental advances enable extremely efficient silo-based hyperscale datacenters that are miniaturized, automated, and unmanned.
At the rack level, the equivalent of 50 to 60 conventional racks of I.T. equipment (at 10 to 12 KW per rack) is miniaturized into the space of about half a rack (roughly 18 cubic feet), which is about 100X smaller.
This I.T. equipment occupies the lower section of a SCRAM Unit (I.T. payload), as depicted here. The upper half of the SCRAM Unit is primarily occupied by infrastructure equipment.
At the datacenter level, "miniaturized" means that a SCRUTINY® silo-based datacenter occupies a volume that is typically 100X smaller than a conventional datacenter of the same power rating.
Moreover, a fully automated underground SCRUTINY® Miniaturized Hyperscale Datacenter has a visible footprint that is typically less than 0.1% of the size of a conventional datacenter (i.e., 1000X smaller) of the same power rating — smaller than a parking space — enabling a massive 4 to 16 megawatt capacity with very little real estate.
The image below depicts the visible footprints of three different hyperscale datacenters, along with their power densities (watts per cubic foot of datacenter volume). Here, the typical entry level for a "hyperscale" datacenter is considered to be thousands of physical servers in a single location. As an example of the high end of "hyperscale," one well-known cloud-based company claims to add thousands of servers daily to one of its larger datacenters.
Each miniaturized hyperscale datacenter is visibly smaller (by 1000X) than a conventional hyperscale datacenter (i.e., less than 0.1% of the visible footprint), but also much more dense (100X denser, at 900 watts per cubic foot vs. 1 to 9 watts per cubic foot for conventional hyperscale datacenters).
Hyperscale providers seeking to reduce risk can easily deploy in multiple locations with increments as small as a half-megawatt (500 KW) each. I.T. equipment can be unloaded, installed, and powered on in less than 10 minutes per megawatt. Infrastructure components (pumps, substation switchgear, transformers, etc.) are deployed in exactly the same way – in under five minutes each – with zero labor.
Each SCRUTINY® miniaturized datacenter can accommodate the just-in-time deployment of I.T. and infrastructure equipment, in any combination from 0.5 MW to 16 MW. If very small capacities are required, any I.T. equipment payload (SCRAM Unit) may be "depopulated" to a little as about 100 KW (0.1 MW), or about 20% of its nominal capacity, as a factory option (it cannot be done in the field).
Innovative datacenter technology that reduces the 10-year TCO by half, for a 16MW hyperscale datacenter. In 10 years, the TCO savings amount to $1 billion or more per datacenter ($100 million per year per datacenter, on average), beginning with a greatly accelerated time-to-revenue and a savings of $80 million in construction costs.
The 10-year TCO savings of $1 billion or more per 16-megawatt datacenter is due to a combination of factors: fastest launch, lowest facility & infrastructure cost, lowest I.T. equipment cost, lowest staffing cost, and an overhead energy and power cost that approaches zero. Although the strategies shown here have the same critical load, the cost of critical load energy and power (light purple area) for the three conventional strategies may appear slightly smaller, because they do not accrue energy and power costs until they launch. On the other hand, due to their delayed launch, they incur an opportunity cost (lost revenue) which is depicted in red.
Datacenter Supply & Demand Drivers
Research firm Jones Lang LaSalle IP, Inc. (JLL), notes that "Today's datacenter landscape is changing fast, as leading providers and users strive to stay competitive amidst rapid regulatory, technological, and environmental change." JLL sums up the economic picture as:
Stratospheric momentum towards the cloud.
A seismic global political shift toward data sovereignty.
New regulations supporting a more sustainable future.
Gartner, Inc. noted that: “More than $1 trillion in IT spending will be directly or indirectly affected by the shift to cloud during the next five years. This will make cloud computing one of the most disruptive forces of IT spending since the early days of the digital age.” [“Gartner Says by 2020 'Cloud Shift' Will Affect More Than $1 Trillion in IT Spending” (Gartner.com, 7-20-2016)]
The individual submarkets and segments are likewise accelerating. According to research by MarketsandMarkets™, "The cloud managed services market size is projected to grow from $27.15 Billion in 2017 to $53.78 Billion by 2022, at an expected CAGR of 14.6%." In a separate report, MarketsandMarkets™ notes that "The modular datacenter market size is expected to grow from $13.07 Billion in 2017 to $46.50 Billion by 2022, at a Compound Annual Growth Rate (CAGR) of 28.90%." According to research by Jones Lang LaSalle IP, Inc. (JLL), "Cloud adoption is racing ahead at an accelerated clip, with the cloud-managed service market expected to double by 2021."
The major cloud providers have a real problem: they cannot build datacenter capacity fast enough, in the right places, to keep up with demand or get ahead of it. The cloud providers have turned to leasing wholesale space from datacenter REITs (Real Estate Investment Trusts) to help close the gap (REITs can speculatively "land bank" and get campuses ready in advance, which speeds up the construction rate). It's been said that half the datacenters that will be needed by 2020 have not even been built yet.
With none of the hyperscale cloud providers having the ability to flexibly deploy hyperscale capacity whenever and wherever it is needed, it is currently a fairly level playing field.
In the trillion-dollar arms race for cloud dominance, none of the major cloud computing competitors appears to have any significant technological advantages over the others – they’re essentially fighting with swords and spears.
The Next Era of Cloud Computing:
Hyperscale Cloud Capacity When and Wherever Needed
SCRUTINY® miniaturized hyperscale datacenters will provide the rapid construction, rapid deployment, capacity, bandwidth, and low latency required to meet future demand – while dramatically accelerating revenue generation and cutting the ten-year TCO in half.
Conventional hyperscale datacenter technology is inherently centralized, costs too much, and takes too long to construct and deploy. Enterprise cloud adoption is already placing accelerating demands on every aspect of datacenter design, even though this acceleration is still far to the left of the “knee” of exponential growth that will be associated with emergent technologies.
Conventional modular approaches cannot provide hyperscale expansion with just-in-time scalability and flexibility. Conventional datacenter technology cannot deal with the explosion of data that will be seen with machine learning, real-time analytics, Internet of Things (IoT), connected cars, and autonomous vehicles.
The current era of huge, centralized datacenters will be superseded by a new era of cloud computing — ubiquitously distributed, miniaturized hyperscale datacenters with dramatically more powerful machine learning and supercomputing capabilities.
SCRUTINY® technology enables future hyperscale datacenters to be miniaturized, decentralized, and ubiquitously distributed — onshore or offshore (underground or underwater).
According to the U.S. Office of Management and Budget, there are 166 combined statistical areas (CSAs) in the United States, comprising 389 metropolitan statistical areas (MSAs) and 540 micropolitan statistical areas — all of which need to be served by local datacenters. Many of these datacenters will need to be modular, just to keep up with growth.
- Michael Thrift, Director
of Data Center Facilities
at Oracle Corp.
Note that the top 55 CSAs each have populations between 1 million and 23.7 million, so such CSAs may need to be served by multiple (or many) local datacenters. SCRUTINY® miniaturized hyperscale datacenters provide the necessary modularity, scalability, and flexibility to keep up with any level of growth from edge to hyperscale, with the lowest possible TCO.
Enterprise transitions to cloud computing -- and emergent technologies, such as AI (autonomous vehicles etc.) are already placing accelerating demands on every aspect of datacenter design. Today’s huge, centralized datacenters will be superseded by ubiquitously distributed, miniaturized hyperscale datacenters. The top 55 of the 166 combined statistical areas (CSAs) each have populations between 1 million and 23.7 million, so such CSAs may need to be served by multiple (or many) local datacenters.
SCRUTINY® miniaturized datacenters provide the necessary modularity, scalability, and flexibility to keep up with any level of growth from edge to hyperscale, with the lowest possible TCO.