by Robert Bryce
“Green” Computing Can’t Power the Cloud
Facebook’s initial public offering was all about superlatives. The May 2012 event was the largest-ever IPO for a US technology company and the third-largest in US history.42 It marked, or so the hype claimed, the coming of age for social media companies. But amid the hype over the company’s stock price, revenues, and growth potential, the media paid almost no attention to the vast quantities of electricity that Facebook and other tech companies need to operate their business.
In April 2012, Greenpeace spotlighted the issue of power demand in data centers in the report “How Clean Is Your Cloud?” The environmental group graded a series of technology companies, including Facebook, Apple, Amazon, and others on the percentage of what it called “dirty energy” being used by their data centers.43 Greenpeace—which, of course, has a Facebook page—gave the social media company a “D” for what it calls “energy transparency.” And the group went on to claim that it had convinced Facebook to “unfriend” coal-fired electricity.44
Never mind that about 40 percent of all global electricity production comes from coal. Let’s consider what the “clean energy” footprint of one of these big data centers might look like.
In 2012, James Hamilton, a vice president and engineer at Amazon Web Services, wrote about Apple’s new iCloud data center in Maiden, North Carolina. Hamilton was responding to Apple’s claim that it was going to use solar energy to help run the site. In a blog posting called “I love solar but . . . ” Hamilton calculated that each square foot of data center space would require about 362 square feet of solar panels.45 In all, Hamilton estimated that powering Apple’s 500,000-square-foot data center would require about 6.5 square miles (16.8 square kilometers) of solar panels. Hamilton said that setting aside that much space, particularly in the densely populated regions where many data centers are built, is “ridiculous on its own” and would be particularly difficult because that land couldn’t have any trees or structures that could cast shadows on the panels.
Utilizing wind energy to fuel data centers would be equally problematic. To demonstrate that, consider the Facebook data center in Prineville, Oregon, which needs 28 megawatts of power.46 The areal power density of wind energy—and it doesn’t matter where you put your wind turbines—is 1 watt per square meter.47 (I will address wind energy in a later chapter.) Therefore, just to fuel the Facebook data center with wind will require about 28 million square meters of land. That’s 28 square kilometers or nearly 11 square miles—about half the size of Manhattan Island, or about eight times the size of New York City’s Central Park.48
The mismatch between the power demands of Big Data and the renewable-energy darlings of the moment are obvious. US data centers are now consuming about 2 percent of domestic electricity. That amounts to about 86 terawatt-hours of electricity per year, or about as much as is consumed by the Czech Republic, a country with ten million residents.49 Put another way, US data centers are consuming about 47 times as much electricity as what was produced by all the solar-energy projects in America in 2011.50
The hard reality is that our iPhone, Droids, laptops, and other digital devices require huge amounts of electricity. According to Jonathan Koomey, a research fellow at Stanford University who has worked on the issue of power consumption in the information technology sector for many years, data centers consume about 1.3 percent of all global electricity.51 That quantity of electricity, about 277 terawatt-hours per year, is nearly the same as what is consumed by Mexico.52 While that’s a lot of energy, Koomey’s estimate of 277 terawatt-hours doesn’t account for the energy used by home computers, TVs, iPads, iPods, video monitors, routers, DVRs, and mobile phones.
In 2013, Mark Mills, a colleague at the Manhattan Institute, wrote a report called “The Cloud Begins with Coal,” which put the total even higher. Mills estimated that when all the energy used for telephony, Internet, data storage, and the manufacturing of information-technology hardware is included, about 7 percent of all global electricity is being used in our effort to stay connected. That amounts to about 1,500 terawatt-hours per year, or nearly as much electricity as is used annually by Japan and Germany combined.53
Regardless of the precise amount of energy being used to run our digital communications network, it’s readily apparent that communications-related electricity demand is growing rapidly. Between 2005 and 2010, global use of electricity in data centers grew by about 56 percent.54 That’s more than three times as fast as the growth in global electricity consumption over that same time frame.55 It’s apparent that the demand for electricity to power data centers will continue to grow as more people, and more things, get connected to the Internet. A plethora of digital devices—ranging from smart phones to GPS-enabled locators on shipping containers—is connecting to the network. In 2012, Intel estimated that there were about 2.5 billion devices connected to the Web. By 2015, it expects there will be fifteen billion Net-connected devices.56 Ericsson predicts fifty billion by 2030.57
Again, the exact numbers are not as important as the trend. The push for Smaller Faster digital devices requires moving ever-more information. The more computing power we use, the more electricity we consume. Big Data has always demanded Big Electron. And as we’ve managed to move more and more bits, we’ve seen a corresponding increase in the demand for electricity.
SMALLER FASTER INC.
INTEL
Official Name Intel Corporation
Website http://www.intel.com
Ownership Publicly traded, NYSE: INTC
Headquarters Santa Clara, CA
Finances Market capitalization: $119 billion58
2012 Revenue $53.3 billion59
Take a moment to look at the period at the end of this sentence. Using current manufacturing techniques, Intel Corporation could fit more than six million transistors onto that little dot.60
Although the vacuum tube ignited the Information Revolution, it took the transistor to make the Information Age relevant to consumers. Transistors are Smaller Faster Lighter Denser Cheaper and more reliable than vacuum tubes. And no other company has had more success at making transistors Smaller than Intel.
Of course, numerous companies compete with Intel, including Advanced Micro Devices and Texas Instruments. But it’s also abundantly clear that Intel, founded in 1968, is leading the world in nanotechnology. By doing so, the company has helped make transistors ubiquitous. Imbedded in hearing aids, smart phones, TVs, dishwashers, computers, automobiles, and dozens of other items, they are among the most commonly manufactured items in the world. Intel alone produces about five billion transistors every second. That works out to about twenty million transistors per year for every resident of the planet.61
Forty Years of Smaller at Intel: From 10,000 Nanometers to 22 Nanometers
Source: Wikipedia. http://en.wikipedia.org/wiki/List_of_Intel_microprocessors.
When John Bardeen, William Shockley, and Walter Brattain assembled the first transistor at Bell Labs in 1947, the device was about the size of the palm of your hand.62 Ever since then, the transistor has been shrinking. Process technology is the manufacturing method used to put transistors and other components on silicon chips.63 In 1971, Intel began producing microprocessors using a 10-micron process technology. (A micron is 1 millionth of a meter, or 1,000 nanometers.) Forty years later, the company was using a 22-nanometer process technology. At 22 nanometers, you could fit more than 4,000 transistors across the width of a human hair. By 2020, analysts believe Intel and other companies may be utilizing a 4-nanometer process technology.64
From 1971 to 2011, the per-transistor price for Intel’s microprocessors dropped by a factor of about 50,000. The company’s modern microprocessors run about 4,000 times as fast as the ones used in 1971, and each transistor consumes about 5,000 times less energy.65 While it’s difficult to imagine those changes, look at the graphic above, which shows how the density of Intel’s microprocessors has increased over the past four decades.
&nbs
p; Forty Years of Denser at Intel: From 2,300 Transistors per Microprocessor to 2.27 Billion
Source: Wikipedia. http://en.wikipedia.org/wiki/List_of_Intel_microprocessors.
In 1971, Intel was able to put 2,300 transistors on a microprocessor. By 2011, the company was installing nearly 2.3 billion on a single chip—a million-fold increase. Back in 1996, Ed Lazowska, chair of the University of Washington’s Computer Science and Engineering Department, estimated that if automobile technology had advanced at the same pace as computer hardware and software, cars would be about the size of toasters and cost just $200. They’d also be able to cruise at 100,000 miles per hour while using just one gallon of fuel per 150,000 miles traveled.66 Remember, Lazowska made that comparison in 1996.
Why pursue such high density? Obvious, says one former Intel guy: “The chip gets smaller because its transistors and wires are smaller. It gets faster because smaller transistors are faster. Smaller is also cheaper . . .”67 The relentless push for Smaller Faster Cheaper has made Intel into a $120 billion company.68
Smaller Denser Cheaper: The Plummeting Cost of Computer Storage, 1956–2010
In 1956, a gigabyte of hard-drive storage cost $10 million.69 By 2010, that same amount of storage could be purchased for about $0.10.70 Today, thanks to Smaller Faster Lighter Denser Cheaper computers and hard drives, consumers can use up to 5 gigabytes (a gigabyte is 1 billion bytes) of online storage—in cloud services like Dropbox and Google Drive—for free. Sources: http://www.geekosystem.com/gigabyte-cost-over-years/; Mat Komorski, http://www.mkomo.com/cost-per-gigabyte.
10
FROM LP TO iPOD
It’s a retro thing to do, no doubt about it. I recently went to Waterloo Records, Austin’s landmark music store, and bought an LP—not a CD, not an MP3, but an old-fashioned 12-inch wide slab of petroleum products. My choice: the twenty-fifth anniversary edition of Paul Simon’s classic album, Graceland. The cost, before tax: $24.99.
I didn’t really need the LP. Shortly before I drove to Waterloo Records, which is about 3 miles from my house, I downloaded the same twenty-fifth anniversary edition of Graceland onto my computer. The download, which cost $14.99 on the iTunes store, took about twelve minutes. It came with seventeen tracks, all of the lyrics to the songs, a digital booklet, a discography of Simon’s albums, and the digital copies of the music videos that Simon produced when the album came out in 1986. In fact, the digital version is, in some ways, cooler than the analog version, even though the LP came with a nice poster.
I dig records. I’ve had a thing for vinyl since I was a teenager and I now have about 250 LPs. CDs are cool, too. I’ve been buying them since about the time Graceland first came out. CDs are Smaller than LPs. Their sound quality is great. They are durable and can handle cold, heat, and other abuse. You can play them at home, in the car, or on the boom box. Those reasons help explain why in the first twenty-five years after CDs became commercially available back in 1982, retailers sold more than 200 billion of them.1
But the LP, CD—and of course, the 8-track and cassette tape—are all anachronisms. In the span of a few decades, we’ve gone from vinyl to CDs and from the Walkman to the iPod. Music that used to exist only when we put the needle of the turntable onto an LP spinning at 33 1/3 revolutions per minute, now lives as a digital file on a machine Smaller than a pack of cigarettes, or even in the cloud as a collection of bits we can retrieve at will from nearly anywhere via services like Pandora and Spotify.
1892: Thomas Edison (center) with his colleagues, gathered around his wax-recording phonograph. The machine could record about two minutes of sound. Source: Library of Congress, LC-USZ62–13413.
That’s a rather long introduction to the obvious: our music has gotten Smaller Faster Lighter Denser Cheaper. Sounds that were formerly available only in the opera house or concert hall at specific times on specific dates can now be heard nearly anywhere, anytime. The whole business of recording, buying, storing, and replaying music has become so easy and efficient that we can now purchase the soundtrack to our lives on our mobile phones without ever setting foot inside the Concertgebouw in Amsterdam or Waterloo Records in Austin.
The push toward iPods and cloud-based music led me to consider my LP collection in terms of weight and density. So I pulled out a scale and a tape measure and tallied the results: my LPs, which hold approximately 2,500 songs, weigh about 62 kilograms and require about 95 liters of space. I then compared that to the music density on an iPod. My findings: an iPod is about 7,000 times as efficient in terms of weight as an LP. In volumetric terms, it’s about 20,000 times as efficient. Put another way, if the iPod Classic—which can hold about 40,000 songs and weighs just 140 grams—had the same density as my collection of 250 LPs, it would be the size of a large refrigerator and weigh as much as a Fiat 500.
Smaller Faster Lighter Denser Cheaper Music Storage: From the LP to the iPod
Over the span of about six decades, our music-playback systems have gotten Smaller Faster Lighter Denser Cheaper. The iPod Classic hold about 40,000 songs and weighs 140 grams—about 5 ounces—making it light enough to strap to your arm or carry in your pocket. If you attempted to carry those same 40,000 songs with you on LPs, you’d need a mule train. That quantity of music-on-LP would weigh about 1,000 kilos, or 2,200 pounds.2 Source: Author calculations, Apple.
Improvements in recording have been under way ever since 1877, when Thomas Edison invented the phonograph. By 1899, when Edison began commercializing the invention for consumer use, the phonograph cost $20 (more than $500 in today’s money) and could record just two minutes of sound.3 Going from Edison’s wax-cylinder phonograph to the iTunes of Steve Jobs took a century, and all along the way, sound recording and playback got progressively better. The wax cylinder gave way to 78-rpm records, which could hold about three minutes of music per side. The LP followed and increased the density of the recording fivefold. The LPs could record about fifteen minutes per vinyl side. The LP gave way to the CD, which prevailed because of its better sound quality, ease of use, and, of course, its higher density. The CD weighed less than half as much as an LP and contained six times as much music per cubic centimeter.4
That ability to produce music, to record sound of whatever type and play it back whenever and wherever we like is an astounding breakthrough. Imagine the expression on Thomas Edison’s face if we were to hand him an iPod and a set of earbuds and then dial up a rendition of Beethoven’s Ninth Symphony. There’s no doubt that Edison would be gobsmacked. And he would be astounded in the same way that his fellow citizens were back in the late 1800s when he began demonstrating and selling his first recording devices.
Today, music and recording technology continues to evolve, making it Smaller Faster Lighter Denser Cheaper than ever before.
11
FROM KUBLAI KHAN TO M-PESA
The SS Gairsoppa was hauling money. To be precise, it was carrying about 219 tons of silver from India to Ireland when it was torpedoed by a German U-boat on February 17, 1941. For seven decades, the wreckage of the coal-fired Gairsoppa and its $230 million cache of silver sat undisturbed on the bottom of the ocean in about 4,700 meters of water.1
In 2011, a company called Odyssey Marine Exploration found the Gairsoppa in a spot about 300 miles off the Irish coast.2 Since 2011, the Florida-based company, which is publicly traded on the NASDAQ under the ticker OMEX and is excavating a number of other shipwrecks and claiming their precious-metal cargoes, has recovered more than 48 tons of silver from the sunken cargo ship.3 The recovery of the silver from the Gairsoppa is one of the deepest and largest recoveries of precious metal in history.4
Many people dream of sunken treasure, pirate booty, and all of the romance that comes with them. But today the idea of using silver or gold to pay debts seems almost quaint. Even good old paper money is starting to appear old-fashioned. Many airlines refuse to accept cash onboard their flights. Instead, customers must pay with credit cards. While cash and credit cards continue to dominate the marketpl
ace, in some parts of the world they are being replaced by “mobile payments,” which rely on the SIM cards and data that sit inside mobile phones.
We are witnessing a move toward Smaller Faster Lighter Cheaper money. While we’re in the early stages of the cash-into-digits trend, the move offers great promise because money fosters interaction. Having a reliable, trustable method of exchanging value—whether with gold, currency, or digits on your phone—builds communities and economies. Access to money is a human right. It allows people to save the fruit of their labor. It fosters the diffusion and accumulation of wealth.
Here’s a brief history of money.
Humans have been using gold as a store of value for millennia. The earliest pure gold coins date to about 560 B.C. in what is now Turkey. The Bible has numerous references to gold. The element appears eight times in Genesis alone.5 Gold coins proliferated for many reasons: the metal is shiny, durable, malleable, easy to test for authenticity, and non-reactive. We can count on it to not dissolve, mildew, or catch fire. Gold was, and still is, pretty scarce. In all of history, the total amount of gold ever mined totals about 165,000 tons, a weight equal to that of one and a half US aircraft carriers.6 Coins made from silver (the word “coin” means “invent”) have many of the same attributes as their gold cousins and have been in use for centuries. While silver and gold—atomic numbers 47 and 79, respectively—have many positive characteristics, they also have a key drawback: they’re heavy.
In the thirteenth century, the Chinese emperor Kublai Khan changed the way we think about money.7 Khan’s great insight was that money—in the form of sea shells or gold coins—was valuable only if people believed in it. He also knew that the different regions of China were issuing their own coins, which made trade within his empire more difficult. So Khan created a new currency based on paper money. By decreeing that his paper money had value, his subjects believed that it did. Khan’s paper money not only provided a common currency for his empire, it was also far superior to gold and silver coins for an obvious reason: it was Lighter. Being Lighter, Khan’s paper money made trade Faster.