AC Versus DC in the Data Center: Redux

February 16, 2012 8 Comments »
AC Versus DC in the Data Center: Redux

What could better fit a month of discussing UPS systems in the data center than a controversial design topic like AC versus DC power? Over the past two weeks, the Data Center Journal has reviewed the basics of UPS systems as well as broad guidelines for selecting an appropriate design. But what if you could use an electrical system that would simplify both your UPS system and other aspects of your electrical infrastructure? Proponents of DC (direct current) power believe this is possible, all with a significant increase in efficiency. Naysayers dismiss these benefits as resting on inaccurate comparisons with AC (alternating current). So, who’s right?

AC Proponents Poisoning the Well?

If the history of science shows anything, it shows that people who are quick to call one thing science and another pseudoscience are just as likely in the wrong as they are in the right. Although science is supposedly neutral, unbiased and based largely on empirical evidence, scientists are often the most dogmatic, biased and committed individuals on the planet. The debate over AC versus DC in the data center—although not as rancorous as, say, the matter of climate change—seems to bring out the dogmatist in many people.

Case in point: in a company blog post, Kevin Brown (Vice President, Data Center Global Offer for Schneider Electric) consistently uses scare quotes around the word “study” when referencing publications on DC efficiency benefits. The tell is that he does this for a paper he admits to not having thoroughly investigated: “We are digging into this one but science is science. . . It’ll be interesting to understand the details of this ‘study’ to see if they tipped the scales in favor of DC.” Even the blog title says much: “Great Hoaxes: Bigfoot, UFO’s, and DC vs. AC efficiency studies.” (One wonders if even the capitalization, or lack thereof, is significant.)

Disagree with the findings of a paper? Fine. But labeling your position as science and the other guy’s position as a hoax, pseudoscience, nonscience or whatever is often just unprofessional.

To be sure, not all AC proponents are dogmatic, and not all DC proponents are detached and objective, and sorting through the hype to find the nuggets of truth can be difficult. But that’s the case for any topic subject to heated debate, whether it’s evolution, climate change or DC power in the data center.

Maybe DC Really Is a Little Better in the Data Center

To his credit, Brown does admit the potential for DC to be more efficient than AC, albeit reluctantly: “there is really very little difference (~2-4% at most) in the efficiency of DC versus a well designed AC system.” So, maybe the arguments for DC power in the data center do have some validity—at least they’re not as unscientific as some might suggest, even if their claims are a little exaggerated. Even if the efficiency improvement isn’t 20–30% but is low single digits, that’s potentially tremendous savings in the long term. And as data centers max out their benefits from virtualization, consolidation and free cooling, they’ll begin looking elsewhere to reduce power consumption: isn’t even a few percent efficiency improvement worth considering?

As always, however, some restraint is needed when facing wild claims about a new or different technology. Every year, some development is touted in the press as the next (fill in the blank), or as something that will revolutionize the (fill in the blank) industry. Most of these do not pan out. Thus, a new strategy or technology should be viewed with caution, and sometimes even skepticism. Nevertheless, one should also keep an open mind.

What DC Has Going for It

Consider the double conversion UPS design: incoming power to the data center is converted from AC (what the electric generator produces) to DC to charge the battery and remove spikes, dips and other power quality anomalies. It must then be converted back to AC for distribution over the typical data center’s infrastructure—which is designed for AC. The conversion back to AC bears some inefficiency, as with any power conversion step. So what’s the advantage of DC in this case? The reconversion to AC can be eliminated, saving some waste energy. Furthermore, since IT equipment generally uses DC (most products convert input AC power to usable DC power internally), another conversion stage could be eliminated: conversion from the “cleaned” AC power (from the UPS/PDU) back to DC once more. The elimination of several power conversion stages is the most widely touted benefit of DC over AC.

Furthermore, eliminating equipment (power conversion stages, for instance) increases available space (a commodity in data centers) and generally improves the reliability of systems (one less thing to break). And it reduces both capital and operating expenses. Some quibbling over other elements of the system, such as reduced losses in cabling and such, takes place as well.

What DC Has Going Against It

Now, the cons. Perhaps the biggest concern for companies building data centers is the lack of equipment designed to run on DC power. The problem here is not so much a technical one as it is a matter of product features. On the other hand, some equipment is designed to use either AC or DC (or just DC), and as more data centers implement DC infrastructure, more products will become available.

Another concern is safety, with some opponents suggesting that high-voltage DC poses a greater risk to data center personnel compared with high-voltage AC (for instance, charge could build in areas of a cable, leading to arcing). The size and weight of cabling for DC relative to AC may also be a concern, although the amount of cabling could be less in a DC design.

But the need for a solid reputation is what really hampers DC. This problem will ease as more data centers adopt DC and as more publications (unbiased and down to earth, preferably) review good DC designs compared with good AC designs, thereby giving companies solid data regarding potential savings for each approach.

And the Winner Is…

Perhaps the biggest impediment to adoption of DC power is the contentious nature of the subject, with each side making claims that either misrepresent the other side or that simply are not accurate regarding its own champion. If some design technique is controversial, a conservative company—and how many are willing to take a big risk in today’s struggling economy?—is likely to pursue the status quo rather than a unique approach.

DC seems to offer some efficiency improvement over AC, and this will no doubt drive increasing adoption of DC power distribution in data centers. Whether it will become a tsunami, however, remains to be seen. Unless vastly superior efficiency can be demonstrated, AC and DC will probably remain competing options rather than one taking the lion’s share of the market (unless AC remains dominant simply owing to historical momentum). In the meantime, a healthy debate is wonderful—but a little respect to and from both sides would be nice.

Author contact

Photo courtesy of Philippe Put.

About Jeff Clark

Jeff Clark is editor for the Data Center Journal. He holds a bachelor’s degree in physics from the University of Richmond, as well as master’s and doctorate degrees in electrical engineering from Virginia Tech. An author and aspiring renaissance man, his interests range from quantum mechanics and processor technology to drawing and philosophy.

8 Comments

  1. Kevin Brown February 17, 2012 at 5:58 pm -

    The majority of my blog is in reference to a study published in 2008 claiming that DC systems are 28% more efficient than AC. I outline 5 specific reasons why that is not true and reference our white paper, published here, and a Green Grid white paper which explain those five reasons in more detail.

    I then comment on a second ‘study’ later in the blog that claims a 15% improvement. I put the word ‘study’ in quotes because the only information I’ve personally seen is a Powerpoint presentation that doesn’t have all the assumptions detailed, so it is difficult to analyze the conclusions. It’s not just that I disagree with the paper, it’s that I have had troubles finding the paper. Hence my use of quotes.

    In your section outlining ‘What DC Has Going for IT’ you fall into the trap of speaking only of conversion steps. It is better to rely on the devices’ published efficiency ratings, which are readily available (especially AC devices). It’s then relatively straightforward to compare the two systems and the differences are small. Why rely on theory when published facts are available? It’s this type of analysis that led me to the UFO/Bigfoot analogy… many proponents draw a vague picture, ignore published data and facts, and then call the other guy dogmatic.

  2. Kevin Brown February 17, 2012 at 7:21 pm -

    Also, you may find the white paper referenced in the first paragraph of my commentary via this link: http://www.apcmedia.com/salestools/NRAN-76TTJY_R2_EN.pdf

  3. Jeff Clark February 17, 2012 at 7:28 pm -

    I understand that you meant to refute–or at least point out a source that addresses–the claims of the study/presentation you referenced. As I said in the article, claims of huge improvements in (whatever) should be viewed with some suspicion. And I don’t think a little sarcasm is necessarily out of order, but perhaps just noting that what some call a study is just a presentation might have been enough (mostly in fairness to the authors, who may or may not have had their work used contrary to their intent)? Just a thought.

    Regarding conversion steps–a conversion is always inefficient (i.e., less than 100% efficient). The point is more that DC may have more potential efficiency than AC in the limit as designs are optimized, assuming the DC design uses fewer conversions. Whether individual products are more or less efficient is not what I was trying to address, although a comparison of products would certainly have some merit.

    Relying on published facts (e.g., about current products) is fine for making a design choice right now. But the article wasn’t intended to only address what’s available now. Published facts AND theory should inform an attempt to anticipate future trends and to determine what type of design has the greater potential. I certainly have no argument with your difficulty accepting the 10%-30% efficiency improvement numbers, and I deliberately avoiding re-presenting those as an argument for DC. But given the energy consumption of data centers, even 1%-2% greater efficiency is valuable.

    As I hinted in the article, a good indicator of which approach is superior will be a comparison of good AC designs and good DC designs, particularly as DC matures a bit. Until then, I’m content to avoid passing judgment on the matter, but it’s certainly fun to watch as the debate progresses.

Add Comment Register



Leave a Reply