Green DC for Datacenters, Why Not Homes Too?
Posted by Bob Warfield on December 19, 2007
There are some posts up about how much greener DC power is for datacenters. Phil Windley says a single server may have a larger carbon footprint than a 15 mpg SUV. That was an amazing statistic for me to hear. CNet reports there is a startup called Validus that is working on the idea of supplying DC within datacenters instead of AC to help combat the waste.
The idea is that converting AC to DC is wasteful of energy, particularly when it is done over and over again inside every digital box. Whether we’re talking wall worts for small boxes or full-on PC power supplies (or worse) for bigger boxes, it is wasteful. I don’t find that hard to believe. All these power supplies get warm to the touch. Heck, my PC supply has a big fan built right into it. This leads me to ask: If this is a good idea for datacenters, why isn’t it a good idea for homes too?
I wonder what percentage of the power we use in our homes would be better off DC rather than AC? Most electronics immediately want to convert to DC. All of our computers and home entertainment electronics are in that category. Many things with motors are not currently in that camp, but they could be with DC motors. I’m speaking of refrigerators, for example. It’s not clear to me that something like an electric range or water heater really cares one way or the other. Certainly anyone that has built a PC can see that it would be easy enough to give up the power supply. The principle issue is simply providing cabling with the right voltages. It’s the same cabling that comes out of your power supply today. One issue would involve those voltages. It’s very easy to change voltages in the AC regime with a simple transformer. DC is more problematic. Unfortunately, even PC’s like several different voltages to be available. A typical PC power supply offers -12V, -5V, 0V (ground), 3.3V, 5V, and 12V. Phew, that would be a lot of outlets on the wall! I wonder what the chances are we could get by on fewer voltage options.
The other interesting thing is that DC is ideally suited for batteries and solar power. As it stands, those who wish to employ solar power have to run the DC output from their solar panels into an inverter to convert it to AC. Of course this costs energy through losses to make the transformation. If we had more of a DC-based economy, life would be better.
So why doesn’t the power company just provide DC instead of AC? There was an epic battle fought between Thomas Edison, a DC proponent, and Nikola Tesla, the father of many early AC inventions, that Edison ultimately lost. It turns out to be cheaper to generate AC with rotating machinery, and cheaper to transmit it over long distances. If we’re going to go DC, we need to do so at the “last mile”. It remains an intriguing idea. I’ll be on the lookout for an “appliance” power supply that will run multiple computers and save power.
5 Responses to “Green DC for Datacenters, Why Not Homes Too?”
You must log in to post a comment.
richnieset said
Please tell me your kidding…
Either they don’t have a physics department at BYU, or Phil Windley is just in to hyperbole. Either way this is the type of statement that puts me over the top. First, let me go on record as saying that I am 100% in favor of reducing energy consumption, regardless of the debate on global warming. AND… I agree with Bob, and everybody, that DC power for the data center is a great idea; one conversion plus less heat equals a no brainer. But I get off the electric bus at the notion that an SUV beats a server on carbon foot print. Let’s coin a new TLA — CFP for Carbon Foot Print.
Ok, now that we have CFPs, the first point I want to make is that this is going to be very difficult to calculate. Let’s consider for example: are we talking lifecycle CFP or just operating CFP. The latter would focus on mainly the energy used to operate the device, and not the CFP associated with acquiring raw material (such as steel mining) manufacturing, shipping to the point of use, and decommissioning. There are so many factors to consider in the lifecycle case, that I think for this response I will confine the analysis to the CFP associated with device operation, rather than the lifecycle CFP.
Now, the average person may be tempted to intuit that the obvious answer is that an SUV has a much higher operating CFP than does a server. And, they would be absolutely right. A little basic math… Let’s assume in favor of the server, that the SUV operator is a “soccer parent” that uses the vehicle only 2 hours per day, for 5 days per week. Since this family takes a two week vacation, that adds up to 500 hours of use per year. Next, we assume the SUV in question has a 250HP V8 (you said it was a gas guzzler); and since the driver is a “soccer parent” they don’t dog it around, so we will assume no more than 33% power usage on average. One other quick fact we need is that 1HP = 750Watts. Based on these assumptions the annual SUV energy consumption is:
750Watts X 250HP X .33 X 500HRS = 31MWh. By the way, these are, assuming no ethanol, virtually 100% generated through the consumption of fossil fuel, i.e. very high CFP. It is at least worth while to note that the CFP of a sever, even on a one to one energy consumption basis, may be lower than the SUV due to the fact that electricity can be produced with a fraction of the CFP per horsepower or KWh (read Solar, Wind, Geo-Thermal, or Nuclear power generation) than that required by a gasoline powered internal combustion engine.
Now for the server. One Dell QUAD CORE server, with 750GB Raid array, requires an 800Watt power supply. Let’s add some extra peripherals to juice up storage and networking and call it a day at 1Kw per hour. So our sever, working at say Google, doing web searches 24/7/365 would consume:
24 X 365 X 1,000 = 8.8 MWh annually. Hmmm…
Bottom line guys and gals is that the SUV that is used, probably less than 15,000 miles per year, requires nearly FOUR times more energy to operate than an IT server in a data center running 24/7/365. Not even close. Not even in the paint!
So my question is this…how many telecommuters can one quad core server support?
smoothspan said
Rich, good for you, run the numbers!
But, maybe it’s not so easy. Let’s try some other variations.
It turns out that the EPA has a calculator for this (your tax dollars and mine at work!): http://www.epa.gov/climatechange/emissions/ind_calculator.html
According to them, the 15mpg SUV doing 15,000 miles a year is responsible for 20,421 lbs of Carbon Dioxide a year (Wow!).
Next, the middle range Dell server (http://www.dell.com/content/products/productdetails.aspx/pedge_6850?c=us&cs=04&l=en&s=bsd) burns 1.5 KW per hour, so that’s 50% more and gets us to 24 x 365 x 1,500 = 13.1 MWh annually. That’s still not where your SUV calculation got us.
If I run this through the EPA calculator, I type in a monthly bill of $109.17 (because that’s how to convert the Dell number to their figure at 10 cents per KWh) and I get back 17,940 pounds annually. Looks like the SUV is ahead, but not by much.
What about cooling that machine? Servers are kept under A/C conditions. Nick Carr’s article via Lawrence Labs (http://www.roughtype.com/archives/2007/02/server_electric.php) claims the ancillary requirements amount to just as much as the server itself. Now we’d be talking nearly 36,000 lbs per year for the server versus a little over 20,000 lbs for the SUV.
Who is right? Hard to say. But if I were looking for issues with the original calculation, I’d look at the 33% load factor on the SUV very carefully. I could not find a quote for an SUV, so I didn’t offer a link, but several web articles claimed 5-10 HP is need to keep a car going 55 mph. All the extra horsepower is only used to accelerate and make the vehicle feel peppy. An SUV will use more, but it isn’t clear to me that the 33% load factor is the right answer.
Or consider another take. Electricity is generated with about 35% efficiency in this country (http://www.ce.cmu.edu/~dh5x/12-090/lectures/elec_guest_lect.doc). That means we need the equivalent of 13.1 MWh / 35% = 37.4 MWh of fuel. Our SUV burns 1000 gallons of gasoline to go 15,000 miles at 15 mpg. That gasoline gallon is good for 60 KWh (http://auto.howstuffworks.com/question424.htm), so 1,000 gallons = 60 MWh versus our server at 37.4 MWh. Now we’re back to the server being ahead, but what about that doubled energy requirement when you factor in cooling, lighting, and so forth?
I do like your thought that the server can use lower carbon footprint power. In the US, we currently generate 28% of our energy through hydroelectric, nuclear, solar and wind (http://www.eia.doe.gov/cneaf/electricity/ipp/ipp_sum.html). Seems to me there is an opportunity to go further.
richnieset said
Bob,
I’m not willing to concede it as a close race too easily, but hey it’s Christmas! I didn’t use the input side of the server power requirements, although I probably should have, since you caught it. Thing is, if you are going to add in the AC, power conversion and lighting (don’t servers work in the dark?), then I think it’s only fair that you add in at least gasoline production, transportation and distribution. We should leave exploration and crude oil production out, seems only fair, since we have to build power plants too. But if we really want to do this right, we should go back to the source.
I am not sure where to start on the energy consumption of a refinery and the transportation, but I’ll give it some thought. I also want to do some checking on your number for the efficiency of electrical production it seems low; but if it is, it’s probably due to the prevalent use of coal and natural gas for power plants. Perhaps we have too many fossil fuel fired power plants anyway. That’s another healthy debate for another time.
As you pointed out, it’s a complex calculation and it is by no means clear that an SUV has a lower carbon foot print than does a server. In my view, a personal SUV at a few hours per week versus a server in today’s cloud computing cluster won’t be able to match up in terms of productivity per pound of carbon consumed or emitted. But one thing is certain; when I am done using my computer, I am turning it off!
smoothspan said
Touche! Good points…
I’m with you: suddenly I’m turning the darned machines off when I go to bed. I’m also wishing Microsoft’s sleep mode was less flakey as that would take the edge off reboot times.
Cheers!
BW
cbmeeks said
One thing you are forgetting about is amperage. Simply supplying enough voltage isn’t enough. You have to be able to “keep up with demand” by making sure a DC power supply can provide enough amperage. Ever wonder why a D cell battery is so much fatter than a AA? That’s why that monster remote controlled truck needs 8 D batteries when your LED penlight only needs 2 AAA. Even though both AAA and D are 1.5 Volts (1.1 – 1.6 average)
The DC PSU for your entire computer would probably be the size of a microwave and it would last 5 minutes with all of your hard drives, dual core CPUs, 4 fans, etc.
Plus, you still need a way to supply power to the DC PSU. Now, if the electric company could provide a DC breaker to your box or if your box could provide an AC/DC converter so that you could plug in DC devices, that would rock. You would have to make sure the power cords were different.
Then, the computer could contain a giant voltage regulator instead of a switching PSU. You could have 12v plugs that could easily be dropped to 5v, 3.3v, etc.
But talk about a major infrastructure change. Oh, and analog TV’s wouldn’t work. But analog is going away one day ( sigh ).
Basically, with that setup, you would be replacing your computer’s switching AC power supply with a big fat capacitor. 🙂
cbmeeks
http://codershangout.com