Power. Electrical power, that is. The very lifeblood of any functioning data center. It doesn't matter if it's the electricity that keeps your server at home running, or the electricity that keeps your employer in business, power is probably the single most important necessity in any IT process. You see, it doesn't matter how well you've developed the application, or how powerful your processing engines are; without electricity to run them, they may as well be boat anchors.

Close examination of how your data center is electrically fed is an absolute necessity. If you're planning on bringing a second data center online, an understanding of how power is produced, transmitted, and distributed is doubly important. Electrical fault tolerance is the end-goal, but without an understanding of how things work, you'll be shooting in the dark. This article examines power from the supply side. It explains how commercial power distribution works, and will help you plan alternate commercial power sources to your data centers. With summer quickly approaching and the demand on commercial power suppliers ever increasing, the possibility of rolling blackouts is, once again, an unfortunate reality. We'll examine power distribution from the local, state, and regional levels, to try and make sense of it all. Hopefully, some of the information here will help you examine your own situation, or ask the right questions to ensure that decision makers and designers are indeed making informed decisions.

The Mechanics of Commercial Power
The Source
Electrical power starts at a generating station, or power plant. Generators within the plant spin and produce electrical current. The generator spins via several means: large water sources (like the hydroelectric stations at Hoover Dam, Grand Coolee Dam, and Niagara Falls), steam (produced via gigantic boilers fueled by coal, oil, gas, or nuclear sources), or gas turbines. Regardless of the process, the end result is to spin the generator, which produces the electricity. Raw power from these commercial generators is typically in the form of three-phase alternating current.

Transmission and Distribution The Transmission Grid
Raw power leaves the generators and enters a transmission substation, the beginning of the transmission grid. The purpose of this grid is simple; it provides the means to transmit power over long distances. Long distances are the first challenge.

Long wires result in power losses. This is a definite problem, unless you plan to transmit power over an extremely short distance. Not too practical? Well then,you need to boost the raw power into a higher range, which allows it to be transmitted further. It's similar to putting a nozzle on the end of a garden hose. The hose takes a certain water volume and pressure from the main. Water hits the nozzle and the resultant pressure is increased due to a smaller opening. The end result is that the water shooting out of the hose now goes much farther than it did before. All the while, the volume and pressure available from the source remain constant. Although this is a simplified explanation (any electrical engineers reading this article, please forgive me), the principles are accurate.

Again, the solution is to boost the voltage to allow it to go further. It's at the transmission substation that the thousands of volts produced by the generators are stepped up to ultra-high voltages (like 155,000 to 750,000 volts), which allow them to effectively be transmitted over long distances via the transmission grid. Electricity in this range is referred to as "high tension." These voltages allow the power to be transmitted for 300 miles or further.

Towers that carry these ultra-high transmission voltages are typically made of steel or concrete. One thing's for sure: they're huge. Three high-tension wires can typically be seen between towers, for the three phases produced by the generators. Ground wires also run between the towers.

The Power Distribution Grid
As most data centers (and toaster ovens) can't use 155,000 volts of electricity, it must be stepped down at some point to a more usable level. Remember, however, that the ultra-high voltage lines won't be seen coming down your street. They'll be distributed to a series of electrical substations. These substations, and related lines leading to homes and industry, make up the Power Distribution Grid.

A simplified explanation of the workings of a substation goes something like this: ultra-high voltage (high tension) is transmitted to the substation. Actually, the transmission grid "drops" power to the substation, and continues to the next and so on. Once inside the substation, transformers step down the voltages to a more usable range. At that point, this lower voltage is sent to a distribution bus, which allows it to leave the substation in a variety of directions and possibly at different operating voltages. For instance, the distribution bus may contain transformers that step down the voltage even further, to 7,200 volts. The distribution bus may also couple the power to voltage regulators to help keep things consistent. This voltage typically makes its way to the individual step-down transformers, which supply power locally (on poles or lawn transformers) to homes and small offices. Power coming into your home is no longer the three-phase high-voltage monster that originated at the power plant.

Remember I told you that an electrical substation steps down voltages, and may also distribute power at different voltages? Well, substations are also connected to, feed, and are fed by, other substations. Higher voltages may also be fed to other locations. This is where the high-grade commercial power used in industrial complexes and office buildings also comes from. In some instances, the substation may actually be built for, or located in, the industrial complex itself. This will satisfy the need for two- and three-phase power at some locations. Higher voltages from these substations are distributed to power-hungry sites, such as large office complexes, malls, factories, and yes, data centers.

The Larger Picture
So, how does this help you? Well, now that you understand the hows and whys associated with power distribution, you can help ensure that your primary data centers are fed via dual feeds, or grids. All this means, really, is that you want at least two electrical sources for your data center. It would be nice if the power originated from different substations if at all possible. If not, you can shoot for separate feeds and separate step-down transformers. But are redundant feeds from the same substation an adequate solution if you need redundant commercial power? Well, not really. I mean,7 think about the last electrical storm that knocked out power in half your town. Chances are that the substation was hit, a switch was tripped, or a transformer failed. Separate substations, if possible, are the better bet. This brings us to what you're really looking for: diversity.

Diversity as a Concept
Diversity is a key element to power distribution. It's the foundation that provides the reliability you seek. Diversity along the local grid exists in several forms. Most notable are the diverse means by which distribution substations receive and distribute power. You see, substations have the means to power adjacent substations. That's right. If a substation loses its high-tension feed, it will continue to receive operating voltages from other adjacent substations. Although it no longer receives ultra-high voltages, the substation still receives higher voltages from surrounding distribution grids. These voltages are high enough to step-down and distribute. So, think of the distribution grid as being meshed at the substation level, because it is.

Diverse routing to your site is an absolute necessity. Just as the substations must be fed via diverse electrical paths, so must your data center. This is a challenge at the local level. Unless the complex where your data center is located is fed via diverse routing (different substations), it's doubtful that diverse feeds to your data center are possible. So this is something to really think about when deciding on a primary or secondary data center location.

Certain buildings in New York City, for example, may actually be fed via four distinct high-voltage feeds, sourcing from diverse substations. Distribution of this type is commonplace in high-occupancy buildings or industrial facilities. Speak to your facilities electricians or representatives from your local utility company for details on how your site is fed. Tell them what you want to do, and work with them toward a viable solution.

Beyond the Primary Site: Secondary Data Centers

Let's say that your firm has decided to build a second data center. The first question to ask is how close the second site will be to the primary data center? If it's in the same general vicinity (like three blocks away), you need to see if it's fed by the same substation. If it is, you'll want to consider an alternate site. Location of your secondary data center is really important from a power distribution perspective. Let's look at power production, transmission, and distribution from three distinct viewpoints: local, state, and regional.

Local Power
We're talking very local here.... Like from a local power plant or large substation to your locations. Substations are everywhere, almost to the point of being so commonplace that you probably don't notice them. So there's a chance that your sites may indeed be fed via alternate power distribution grids. Source the substation to the next transmission point up the line. Is it fed from the transmission substation at the power plant? If so, this is a good thing. A single power plant feeds many towns. Just ensure that your sites are (or will be) fed from different substations. Chances are pretty good that they are.

But will it be good enough? That depends on your firm's individual goals for electrical fault-tolerance. Remember that most substations are meshed with the next closest substation. You made decide that you need to have the secondary site fed from a different power plant.

Statewide Power and Distribution
No power plant is an island. They feed, and are fed by, other power-producing plants within your state. Just as there is a local power transmission and distribution grid, there is a state-wide grid as well. Within that grid is a network of interconnected power plants that help to ensure power demands are met for the entire state. If the production capacity of a local plant is deficient at peak times, the statewide power grid augments the available supply of power to the area. Building additional power-generating plants is the best way to ensure that supply meets or exceeds the demand. But the construction of new powerhouses hasn't always kept up with the total demand for power. So the notion of regional grids has reached the forefront.

The Regional Approach
California recently found itself in a power crunch. New York may soon be heading in the same direction. In these instances, the demand for power is outstripping the production capacity of the statewide power grid. So, what can be done to keep up with demand? Building additional power plants is the correct solution, but who wants one of these facilities in their backyard? These are legitimate concerns, and the end result is the delay of progress in any planning, permits, or new construction. For these reasons, states are looking to their neighbors for help.

Prodded by the federal government, action has been taken toward the creation of regional power grids. Hypo-thetically, all northeastern states, for example, may join (or be forced to join) an "official" regionalized power grid. Although a good idea, it has its own set of problems - rate control; local and state environmental authority; and the long-term role of state-sponsored public utility authorities that typically control the electric companies.

None of these obstacles are insurmountable. In actuality, the "grid" extends far beyond statewide boundaries already. Metering, economics, and regulation can sometimes be more complex than the technology needed for the actual generation and distribution of power, but these issues can and will be worked out. Whether it's for the better remains to be seen. The bottom line is that certain states are heading for an energy crunch and something must be done.

Federal Dams and Power Marketing Administrators
Believe it or not, the federal government is in the power business. Huge hydroelectric power plants at the Grand Coolee and Hoover Dams are examples of the megawatts that these facilities generate. Transmission lines from these facilities provide power to a number of locales. In the case of Grand Coolee, this includes portions of Northern California. In this scenario, generators are phased in and out based on the load. Even with this power augmentation, California is still in a crunch. For this reason, with power being brought in using "regionalized" methodologies, a best bet may still be to build in another geographical location.

This is why an understanding of the production and distribution of commercial power is so very important for the IT professional. Asking the right questions can save you a lot of pain in the future, especially if your data centers aren't backed up by generators or a UPS capable of sustaining operations for an extended period of time.

BrownOuts
Worse than a blackout is the infamous brownout. Sometimes there simply isn't enough electricity to go around. When this happens, unless power consumption is quickly brought under control, insufficient levels are distributed along the local grid. Remember that garden hose analogy? Now imagine trying to feed 100 lawn sprinklers with that same hose. Doesn't work too well, does it?
The term brownout was coined in the 1960s, named for its effect on incandescent lighting (bulbs burned dimmer, or brown, instead of bright white). Now a lightbulb may not care what voltage you feed it, but sophisticated equipment does, especially computers.
Computer power supplies may come with over-voltage protection (within reason), but consistently feed it with insufficient power and watch out! Everything from logic and memory processes to how fast the cooling fans and disk drives spin is in peril. And, while we're talking about cooling functions, think about how that power-starved motor, turning the data center's air conditioning compressor, feels. Brownouts have long-reaching effects. It may not affect your data center today, but it will shorten mean-time between failures of components, void manufacturer warranties, decrease the availability of common repair parts in a region (from all the repairs on everyone else's equipment), and who knows what else!
To prevent the damage associated with brownouts, power companies control power demand through a rolling blackout. A blackout occurs when power is absent from a location - your house, neighborhood, town, and so on. When the power grid can no longer keep up with demand, emergency procedures must be enacted to immediately reduce system drain. The resulting action is to shut off power to "pockets" at the source. Since no single area can be affected indefinitely, each "takes turns" during the day. The end result is a "rolling" blackout. While these blackouts may help the power authority, they can be devastating to local businesses. Think of yourself as an ice cream shop owner without power to your refrigerators for five hours a day, or a sales rep who can't process an order!

Build Further Away
Regional power grids may indeed help the situation, but may not be the "belt and suspenders" you're looking for. If your company has the real estate in a different region, you may consider building your secondary data center there. You'll be assured that power is supplied via different local, state, and regional power grids. The data center will also be insulated from the possibility of city, state, or regional disaster, either natural or man-made.

Once power has reached the data centers, it's important to decide, and design, effective distribution internally (e.g., within the data center itself). This is where the notion of harmonics, power balancing, load distribution, UPS, and generators come in. If you want to help ensure power continuity to mission-critical equipment, getting the power to your site is the first step, but a huge one!

Conclusion
Now that you understand how power is produced and distributed, you can see the importance of comprehensive planning in data center design and construction. Proper design can save your job, and may someday save your company (literally). If your firm depends on its computing environment, a rolling blackout or any extended power outage can be devastating to your business. Keep this in mind. It's going to be a long, hot summer...