Cooling Data Centers Could Prevent Massive Electrical Waste
The energy used by the nation’s servers and data centers is growing at an unsustainable rate. Not only that, but web servers are notoriously inefficient. For example, computer servers are used at only 6 percent of their capacity on average, while data center facilities operate at roughly 65% to 75% efficiency, meaning that 25% to 35% of all the energy consumed by servers is wasted (converted to heat).
If we are to even consider reducing our energy consumption and carbon footprint, the growing demands generated by our web servers must be near the top of the list of possible improvements. And the Department of Energy agrees.
Researchers at DOE’s Pacific Northwest National Laboratory (PNNL) in Washington and National Renewable Energy Laboratory (NREL) in Colorado are hard at work figuring out ways to make our data storage infrastructure more efficient by running them at lower temperatures. The technology exists to achieve efficiencies of 80% to 90% in conventional server power supplies. Moving this heat source away from the server allows the cooling efforts to be focused on the computing elements.
Alternative Cooling Approaches
(from PNNL’s Energy Smart Data Center)
- Evolutionary progress is being made with conventional air cooling techniques that are known for their reliability. Current investigation focuses on novel heat sinks and fan technologies with the aim to improve contact surface, conductivity, and heat transfer parameters.
- One of the most effective air cooling options is Air Jet Impingement. The design and manufacturing of nozzles and manifolds for jet impingement is relatively simple.
- The same benefits that apply to Air Jet Impingement are exhibited in Liquid Impingement technologies. In addition, liquid cooling offers higher heat transfer coefficients as a tradeoff for higher design and operation complexity.
- One of the most interesting liquid cooling technologies are microchannel heat sinks in conjunction with micropumps because the channels can be manufactured in the micrometer range with the same process technologies used for electronic devices.
- Liquid metal cooling, used in cooling reactors, is starting to be an interesting alternative for high-power-density micro devices. Large heat transfer coefficients are achieved by circulating the liquid with hydroelectric or hydromagnetic pumps. The pumping circuit is reliable because no moving parts, except for the liquid itself, are involved in the cooling process. Heat transfer efficiency is also increased by high conductivity. The low heat capacity of metals leads to less stringent requirements for heat exchangers.
- Heat extraction with liquids can be increased by several orders of magnitude by exploiting phase changes. Heat pipes and Thermosyphons exploit the high latent heat of vaporization to remove large quantities of heat from the evaporator section. The circuits are closed by either capillary action in the case of heat pipes or gravity in the case of Thermosyphons. These devices are therefore very efficient but are limited in their temperature range and heat flux capabilities.
- Thermoelectric Coolers have the ability to provide localized spot cooling, an important capability in modern processor design. Research in this area focuses on improving materials and distributing control of TEC arrays such that the efficiency over the whole chip improves.
Capturing Waste Heat
Reusing the waste heat from a data center may not make the server room itself more efficient, but depending on how heat is reused, it can save a company a significant sum of money. In its report to Congress last year on data center energy consumption, the federal Environmental Protection Agency suggested the practice. And the idea has gained traction, according to Mark Fontecchio of SearchDataCenter.com.
For example, in Winnipeg, Canada a media company called, Quebecor, efforts have been made to take the heat from the 2,500-square-foot data center on the ground floor and use it to heat other parts of the building.
Because of the cool Winnipeg climate, engineers decided to make use of that cool air by installing air-side economizers that draw in outside air. The economizers include baffles that open to varying degrees depending on the outside temperature and how much cooling the data center needs.
After the air cycles through the approximately 100 eight-way servers, it warms up in the process. It then goes into an overhead plenum, where about 10% of the air is re-circulated to warm the outside air that comes into the data center.
Another duct out of the exhaust plenum to the intake duct of the editorial office upstairs. Quebecor also added a second thermostat to its editorial offices; the first controls the traditional heating furnaces. That whole process used up another 60% of the waste heat. The data center dumps the remaining 30% into the adjacent warehouse.
AC-DC?
For electricity flowing all the way from power plants to the wall socket, alternating current is far superior. But for the short transmissions inside those computers DC power prevails. The search for ways to convert AC to DC more efficiently is leading some data center companies to consider a DC-centric approach.
It’s easier to transmit AC over long distances; DC requires thick copper cables or bars, instead of comparatively lightweight wires. But DC becomes a more serious possibility for power once AC reaches a building.
Converting from one form of power to another in a computing environment may not be performed efficiently, especially at the server level, and even then, the resulting waste heat may be deposited in the rack or computer room at a point that requires further effort to dispose of it with the air handlers. Unfortunately, there is disagreement in the community over how to address these inefficiencies.
- DC advocates argue that plugging servers into AC power is inefficient, and switching systems to DC would cut down on waste heat and component failure.
- Proponents argue that using DC outside the server removes some of the inefficiencies of power supplies that convert AC electricity to DC. Servers without such power supplies don’t have to contend with as much waste heat and attendant component failure.
But according to NPPL, substituting DC power in data centers as a replacement for conventional AC power has not yet made significant inroads into many data centers because the technology is unfamiliar to many facility engineers.
Despite the wide-spread use of DC power in telecommunications, there is reluctance within the computer industry to switch to new technologies without field experience showing that the switch could be done safely and would have operational and economic benefits without causing unanticipated problems.
If DC would in fact be a more efficient type of power within servers themselves, might it be possible to site server farms to take advantage of the DC provided by integrated renewable energy generating systems such as solar PV and wind?
Other Posts on Data Centers:
“Is Paperless Really So Green”:: Green Building Elements
“Don’t Forget to Turn Off the Servers”:: EcoPreneurist
Photos:1.© Eimantas Buzas | Dreamstime.com; 2. Versageek via Flickr under a Creative Commons License
timmy - you’re retarded!!! do some research man!?
YOU’RE IDEA OF “ALL DC POWER” IS COMPLETELY IMAGINARY… DO YOU EXPECT TO REPLACE EVERY ELECTRONIC DEVICE IN EVERY COMMERCIAL AND RESIDENTIAL BUILDING!? DC POWER IS ONLY USEFUL (ECONOMICAL) IN HVDC TRANSMISSION LINES, NOT DISTRIBUTION LINES. “AC OR DC?” WAS DECIDED AND PROVEN HUNDREDS OF YEARS AGO BETWEEN EDISON AND TESLA!
jweezy-
Really? Retarded? Ouch.
I strongly suggest you re-read the post. I would especially urge you to find the portion of the text where I propose the idea of “ALL DC POWER” [your quotes]. If you are able to find that, let me know and I will buy you a beer.
At no point in the post do I suggest we convert to all DC power and replace every electronic device in every commercial and residential building.
I mean no disrespect, but I honestly have no idea what you are talking about.
It’s important to realize that data centers aren’t composed of top-secret, specialized supercomputers. They’re built from the same servers, routers, disk drives, etc. that are used by businesses and individuals every day…just a whole lot more of them. These components are mass produced for the marketplace. It would be impractical to build components that operate solely on DC power when there is no customer base, and impractical to convert a data center to DC power distribution when there is no equipment available to take advantage of it…simple chicken-and-egg situation.
Yes, there is a lot of room for improvement in efficiency of computing equipment. However, it’s not fair to single-out data centers as wasteful power hogs. Don’t forget about the millions of personal computers in homes and offices around the world. Every one of them has a little box inside that converts AC to DC (quite inefficiently, if you believe the article) dumping waste heat in the process. And every one of them has a processor chipset inside which also wastes energy. Time and effort are better spent in improving the efficiency of the components which make up the data centers than figuring out more elaborate ways to cool an inefficient design. Incidentally, equipment manufacturers are already doing this…multi-core processors that operate at lower speeds, for instance.
Average server utilization is also an irrelevent statistic. Yes, on average, a server may use only 6-percent of it’s capacity. However, these systems must be over-designed to handle peak load for the simple reason that Mr. Websurfer won’t tolerate having to wait more than five seconds for his MySpace page to load. For certain, these servers operate near full utilization at peak times of the day, even if they are lightly loaded at other times. Data centers are designed as they are because of the demands we place on our IT infra-structure. We want it cheap, fast, and reliable…it’s hard to get all that and high-efficiency too.
There are already many servers available with DC power. You can set up today an enterprise-class datacenter with all DC powered components, from servers to switches to routers to storage networks. IBM, Sun, and I’m sure many others have commodity servers with DC options. See IBM X3650 T for an example.
DC Power…. Ok People… what 99% of you dont realize is that the components INSIDE Your computers (desktops, laptops, servers, whatever ARE DC… thats the whole point of the power supply.. Take 110/220 VAC in and convert it to a coupleof different DC voltages…
Now, having used DC power for years in Telecom and Telecom related environments I dont see how people can say there is not expertise out there for this stuff in the DataCenter… one just has to visit a place Like Level3’s Courtland St Data Center in Atlanta to see that it is possible…
Not only is DC possible but it reduces the complexities of Redundant Power via elimination of costly and inefficient DC to AC Inverters…
“The energy used by the nation’s servers and data centers is growing at an unsustainable rate. Not only that, but web servers are notoriously inefficient. For example, computer servers are used at only 6 percent of their capacity on average”
I believe in the future data centers will deploy more of a virtual private system approach on large scale servers. It would greatly improve efficiency.
Telco operations had been all DC for quite a while.
Essentially telephone switching centers running at 48V DC were the original computer data centers.
With the switch to cheaper generic components Telcos have relaxed their requirement for DC systems and have migrated over the years to AC.
It would be interesting to see the world switch back to the Telco 48V standard!
Many data centers are DC supplied by replacing the 120volt power supplies with 48 volt models that connect to a DC bus bar in the system’s rack. This is considerably more efficient and also allows for the heat generated by the power supply to be contained separately from the rest of the hardware.
You are probably connecting for the internet through hardware running on 48 volt DC power. Phone and cable companies almost exclusively use these types of systems and have you looked into Cisco equipment lately? You can order almost any router from them with a DC power supply instead of AC. So please, don’t post your stupidity unless you have some idea of what you are talking about.
Mr. Sinister and Jweezy — I think you doods missed the point about AC/DC. The author points out that the components run on DC already. Why have redundant power supplies in each server when the data center could theoretically switch the power for all devices centrally? Consider one big power supply (and of course a backup) for all devices instead of hundreds or thousands of small power supplies in each chassis.
http://en.wikipedia.org/wiki/Economies_of_scale
Seems to me a WHOLE LOT of context is missing from this article. Going to a DC power bus for distribution to servers and networking gear means the data center’s UPS doesn’t need an inverter on its backside. And the individual pieces of gear (servers, switches) don’t need a rectifier on their front side of their switching power supplies. That’s all. And if it were a relatively high voltage bus, say 350 VDC, the busses and power cables would need 1/3 the copper of standard 115 VAC, again not a biggie in efficienty. A new industry standard, with its own style power connectors and with a whole lot more careful attention to polarity. Plug in your server to this power bus backwards and it’s toast, not so with an AC power bus. When it comes to heat production in data centers, rectification (AC to DC, in each power supply) and inversion (DC to AC, in the continuously online UPS) aren’t the first things that come to mind… unless one makes power distribution equipment for a living and does not make servers or networking gear. The telcos have used 48 VDC for years, but that’s because the original exchanges really did run on batteries, 24 lead-acid cells in series. When it comes to wasted energy, a lost Watt-hour is a lost Watt-hour, whether it originally came from DC or AC.