When the Steel ORCA team started planning, it was our intention of being the “Greenest Data Center on Earth,” and, as a data center/services company, we were in the enviable position of being unrestrained by established corporate policies, legacy designs, or outdated facilities. Beginning with a completely blank slate, we embarked on a two-year effort of looking at virtually every advanced design, concept, and product that showed promise of delivering real energy efficiency while maintaining or improving reliability and offering reasonable economic return.

Along the way we found numerous solutions and were constantly challenged by new ideas/products in the development pipeline that outperformed the latest advanced products in the marketplace. It quickly became apparent that if we were to lay claim title to being “Greenest,” we would need to look towards advanced technologies designed to support the next generations of technology as well as capable of sustaining the legacy technology today. High-voltage direct current (HVDC) is one of the technologies we examined, and it became our preferred power distribution medium.

At the start, we were wary of all the vendor hype around the application of HVDC in a data center and shared all the concerns of the numerous naysayers. We participated in online discussion groups where many knowledgeable participants wrote off HVDC as a fad, as a dangerous design, and as a solution without products to support it; they offered numerous other reasons to just to say it was a bad idea. Yet, as we did more research and connected to people who had implemented HVDC and identified vendors producing the products from high-quality rectifiers to the last 10-feet of connectivity (cords, plugs, outlets, and server power supplies), we slowly became believers that the market is ready for large-scale 380-Vdc implementation.

HISTORY

Interest in HVDC started to develop from 2004 through 2006 with work at EPRI and proof of concepts (POC) executed at Sun Microsystems; however, manufacturers needed to develop standards for connector types and specifications for the dc power supplies and rack-mounted power strips. Concurrent with the technology connectivity and product development, the electrical design community needed to see several POC installations to fully evaluate the impact to the data center support infrastructure. One of the first facility demonstration projects to gain recognition was at Syracuse University in 2009, where Validus (now part of ABB www.validusdc.com) installed a HVDC system to power an IBM Z10 mainframe.

Then in 2010, IBM announced two servers running on 380 Vdc, one POC at the University of California and the other POC at Duke Energy. You can watch the online video demonstration of a live direct current environment at http://hightech.lbl.gov/dc-powering/videos.html.

Then came 2011, and connectivity products started making it to market, with companies such as Universal Electric (www.uecorp.com) launching Vdc busway products and HP modifying its power supplies to be hot swappable. These changes enable ease of conversion to a HVDC environment.

One of the surprises we found in our research was a NTT R&D analysis that identified numerous facilities in Japan, Europe, and the U.S. that currently are using HVDC.

With a reasonable installed base, 380 Vdc is rapidly moving from bleeding edge to leading edge, and with so many suppliers, manufacturers, and technology suppliers supporting 380 Vdc our interest was reinforced. Yet as  operators, we knew that if we became a 380 Vdc shop we would need to provide our clients with a migration path to convert over time.

CATALYSTS BEHIND HVDC

An overall energy-efficiency improvement between 8 and 15 percent is an obvious driver, but there are also other significant attributes that should not be overshadowed by the energy calculations. These include space, reliability, CapEx savings, reduced carbon footprint, and safety.

Experts can debate the size of efficiency improvement, but suffice it to say that removing multiple power conversions in the power train will produce savings. At a minimum the HVDC design removes the inverter from the UPS and eliminates the first ac/dc conversion in the server power supply. Transformer PDUs sitting on the data center floor can also eliminated. So the savings are real and not vapor watts.

Space and reliability are directly related to the elimination of equipment and components. Less equipment equals less space, and fewer components statistically equals improved reliability. CAPEX and OPEX savings follow the use of less space, less power, and therefore less cooling. All of the above factors contribute to lowering the carbon footprint to build and operate.

Then there is the safety issue. This is probably one of the least understood functions in using HVDC. One member in a discussion group even went so far as to say that he would rather be shocked by ac than dc while acknowledging that either would be fatal.

Such statements were typical of comments from people with limited to no direct experience in the HVDC topology.  Because I do not consider myself an expert in short-circuit calculations, I sought out industry experts to find that a properly designed HVDC application can deliver significantly lower short circuit currents than its equivalent Vac counterpart.

WHY 380 VDC?

There have been numerous discussions on this matter as the dc power supplies in servers operate across a wide range of dc voltages. The power design inside the servers goes something like this:

The server receives ac power from an outlet, corrects it for power factor, and converts it to +380 Vdc, and then back to ac (double conversion) before making a final conversion back to dc and a breakdown into the multiple low-voltage dc levels that the electronics actually operate at.  

By eliminating the initial power factor correction and ac/dc conversion we eliminate conversion losses in the server. Although this will increase a data center’s power usage effectiveness (PUE), it should at least be partially offset by the facility energy savings through the elimination of the UPS inverter and a reduced cooling requirement. The focus is not PUE but achieving the lowest overall energy consumption.  The server power supplies are generally set at +380 Vdc so this becomes the logical HVDC voltage. There are parts of the globe however where the focus is on other HVDC ratings (240 Vdc).

So back to the title, “The Time Has Come for 380 Vdc.” We have done the research on 380 Vdc along with a wealth of many other promising technologies that are hitting the market. Now it is time to develop the implementation strategies, including providing a migration path for our clients. Both the engineering details and the transition planning are proceeding. In the coming year you will see our plans develop along with many others pursuing the benefits of 380 Vdc. This is a solution to watch, as it will continue to pick up market share. 

This is just one of over a half dozen new solutions we have selected to implement in order to gain maximum efficiency in our operations for years to come. The process is our Kaizen (translation kai (change), zen (good) is improvement). This method became famous by the book by Masaaki Imai titled Kaizen: The Key to Japan’s Competitive Success.

The core principle is the (self) reflection of processes (feedback).

The purpose is the identification, reduction, and elimination of suboptimal processes (efficiency).

 We encourage all data center operators to establish their own Kaizen or Continuous Improvement Process (CIP). It is a passion. We owe it to our clients, to our society, and ourselves.