Academic Integrity: tutoring, explanations, and feedback — we don’t complete graded work or submit on a student’s behalf.

SUMMARY Google operates some of the largest data centers in the world to support

ID: 413017 • Letter: S

Question

SUMMARY

Google operates some of the largest data centers in the world to support its search engine,

?mail, Google Maps, YouTube, and a host of other applications delivered to a global

audience. Google’s Technical Program Manager describes how Google manages the energy

consumption of its data centers using industry best practices. L=10:01.

CASE

Consumers of technology constantly demand devices that are smaller, more efficient, and

more powerful than the ones they have. But most consumers don’t understand the massive

back-end infrastructure that powers their “front end” devices, like mobile phones, smart

-

phones, tablets, and desktop computers, nor the impact that this infrastructure has on the

environment.

Take, for example, the smartphone. iPhones, Androids, and BlackBerrys represent a trend in

all forms of mobile technology towards smaller devices that perform an increasingly large

number of functions. But every time a smartphone user connects to the Internet, places a

call, or sends an instant message, it uses power not only on their phone, but at every step

of the infrastructure used to perform that function. More often than not, data centers, also

called “server farms,” are intimately involved in any Internet-based communication. These

data centers are growing not only in number, but also in sheer size. IBM has a data center

which covers approximately forty thousand square feet (three football fields) and contains

ten thousand servers. Major data centers require such a large amount of energy to power

Chapter 1,

Case 2

GooGle Data Center effiCienCy Best praCtiCes 2

continued

and maintain that only large corporations are able to build them up. The cost of running

large data centers is a significant component of the overall IT budget of firms. There are two

components to the energy cost of data centers: the cost of running the computers and the

cost of cooling them. In the large data centers, temperatures would reach 120 degrees in

just two minutes if the cooling system were disabled. At this temperature, processors and

hard drives begin to malfunction.

In 2016, there are approximately 3 million data centers in the United States and over 9

million data centers of all sizes worldwide. The number of data centers is growing at

around 15% in 2015. IDC estimates that the number of data centers will decline beginning

in 2018 due to the growth of very large cloud mega data centers. Currently US data centers

consume about 100 billion kilowatt hours of electricity. Most data centers do not practice

energy management, and waste approximately 20 billion kilowatt hours annually. The

growth of cloud computing, streaming of music and video, is expected to accelerate data

center power consumption in the next ten years even as the number of data centers decline.

Due to these increasing power demands, by 2020, the world’s computer servers will match

or exceed the carbon emissions of the airline industry. The fastest expansion in data centers

is now occurring in developing markets like those of China, Brazil, and Argentina. Making

computers greener won’t just be environmentally beneficial. It will also relieve the financial

burdens of companies that maintain these server farms. There is plenty of incentive for

online e-commerce companies like Amazon, Apple, Google, Facebook, and Microsoft to

reduce their data center power consumption. In addition, competition among the industry’s

largest providers of data centers (IBM, Oracle, HP, and Intel) are incentivized to help their

customers “go green” and reduce costs. The industries largest providers of cloud services,

such as Amazon, also have a strong incentive to reduce power consumption.

One method that companies use to conserve energy is to distribute their data centers

worldwide, placing them in areas where it’s more efficient to power them. For example,

placing data centers in areas of the world where ambient temperatures are lower lessens

cooling requirements. An area in high demand is Iceland, and other Scandinavian countries,

where the temperature is much lower and cooling needs are reduced. In addition these

countries make extensive use of hydropower, which does not contribute to global warming

and is much less expensive than coal and nuclear power. In the United States, the hydro

-

power produced by the Columbia River makes the northwest a popular data center location,

along with southeastern states that rely on hydropower produced by the Tennessee Valley

Authority and the Tennessee River.

Another method for reducing the IT-component of power consumption is virtualization.

This technique allows servers to perform multiple tasks at once and is one of the more

prominent “green” initiatives for reducing emissions and increasing efficiency. Using this

Chapter 1,continued

technique, a single machine can run more than one operating system at the same time,

and operate at a much higher duty cycle, reducing the number of computers required to

perform the same number of tasks, and reducing the overall cooling requirements of data

centers. On the other hand, increasing the duty cycle (the amount of time a computer is

actually doing work rather than idling) also increases the heat generated by processors.

Google is considered to be an industry leader in data center efficiency, in part because it

builds its own custom data centers rather than rely on standard industry equipment and

practices. However, although “going green” can be both financially and environmentally

beneficial, companies that develop unique technology are given an incentive not to share

them to gain a competitive advantage. For example, Google has proprietary virtualization

technology that it won’t share, and is secretive concerning the number of servers it oper

-

ates. However, as demonstrated in this video, Google takes many other steps to share its

energy efficient best practices with other firms and the public.

1. What is PUE, and why is it an important place to start when considering how to reduce

data center power consumption? What value of PUE should data center managers strive

f o r?

2. Describe the five methods recommended by Google for reducing power consumption.

3. How much of the world’s global greenhouse gases are the result of computing? List

several factors that will tend to retard or accelerate data center power consumption both

globally and in the United States.

4. Where do you suspect that data-center power consumption will be greatest: developing

and emerging economies or already developed economies? Why?

5. Imagine that a company has developed an advanced technology that allows it to reduce

its data-center requirements by an unprecedented amount, and creates a competitive

advantage for the company in the data-center market. Why should it share that technol

-

ogy with other data-center firms? If this firm does not share its techniques, the rest of the

industry will continue to operate less efficient centers, and increase global emissions of

greenhouse gases above what they would otherwise be.

6. Should the government or an industry association regulate the carbon emissions of the

data center industry as they do the airline industry? Or the automobile industry? Or is it

sufficient to rely on the competitive market place to drive down energy consumption in

data centers?

3.Where do you suspect that data-center power consumption will be greatest: developing

and emerging economies or already developed economies? Why?

4.Should the government or an industry association regulate the carbon emissions of the

data center industry as they do the airline industry? Or the automobile industry? Or is it

sufficient to rely on the competitive market place to drive down energy consumption in

data centers?

Explanation / Answer

a). Measure PUE

b). Manage airflow

c). Adjust thermostat

d). Use free cooling

e). Optimize power distribution

3. Around 2% of the global greenhouse gas emissions are a result of computing. In this 2%, 15% constitutes data area and 5 % includes large internet data area.

Factors that accelerate power consumption

1. Advancements in IT.

2. Cloud computing.

3. Audio and Video streaming.

4. a). Data area power consumption will be high in developed countries, because most of the major data areas are located in these countries.

b). Now, to reduce data area power consumption, companies are preferring places with lower ambient temperatures. This includes Iceland and Scandinavia.

5. Obviously if a company has found an advantage over the competition they would be reluctant to share it because they can outperform their counterparts. We believe that if a company had such an advantage, they would only share it based on a “greater good” concept. Meaning, more efficiency in the industry and a reduction in greenhouse gasses is good for everyone. Perhaps marketing this advanced energy saving technology as a service would still yield a competitive advantage and also reduce greenhouse gasses.

6. a). A government or an industry association should regulate the carbon emissions of the data area industry as they do for other industries.

b). Such an association can make the companies aware about the importance of „going green and thereby reducing emissions.

c). c). In addition, competition among the industry’s largest providers of data area are incentivized to help their customers “go green” and reduce costs.