STUDY OF ENERGY EFFICIENCY ON PORTABLE DEVICES USING CLOUD
COMPUTING: THE CASE OF OFFICE PRODUCTIVITY APPLICATIONS
A Thesis by
Hemanth J. S. Kothuru
Bachelor’s of Technology, Vellore Institute of Technology University, 2007
Submitted to the Department of Electrical Engineering and Computer Science
and the faculty of the Graduate School of
Wichita State University
in partial fulfillment of
the requirements for the degree of
Master of Science
December 2010
© Copyright 2010 by Hemanth J. S. Kothuru
All Rights Reserved
STUDY OF ENERGY EFFICIENCY ON PORTABLE DEVICES USING CLOUD
COMPUTING: THE CASE OF OFFICE PRODUCTIVITY APPLICATIONS
The following faculty members have examined the final copy of this thesis for form and content,
and recommend that it be accepted in partial fulfillment of the requirement for the degree of
Master of Science with a major in Electrical Engineering.
__________________________________
Vinod Namboodiri, Committee Chair
___________________________________
Preethika Kumar, Committee Member
___________________________________
Ramazan Asmatulu, Committee Member
iii
DEDICATION
To my loving parents and sister for their endless support and patience;
and to my admirable teachers for their priceless knowledge and grateful support
iv
ACKNOWLEDGEMENTS
I am extremely thankful to my admirable thesis advisor, Dr. Vinod Namboodiri, for his
dedicated support, encouragement, and supervision. He always made time for me in spite of his
busy schedule and offered me helpful guidance in my academics, career, and life. I also take this
opportunity to thank members of my committee for their time and effort.
I would like to extend my gratitude to Dr. Ravi Pendse for his valuable advice and
suggestions throughout my career at Wichita State University. It has been a privilege to work
under him as a graduate research assistant.
I take enormous pleasure in recognizing, with endless thanks, all who provided me with
laptops and all who helped me in numerous other ways. Without them, this thesis would not have
been possible.
v
ABSTRACT
Today, there is an exponential growth in the use of laptops for computing and
communication. However, the battery life of laptops is only a few hours, at best. Furthermore,
studies indicate that laptops contribute to approximately 1% of the overall global energy
consumption. Thus, there are significant incentives to minimize the energy consumed by laptops.
To achieve this goal, it is important to understand the energy expended by each component of a
laptop.
Initially in this work, the power consumed by each component of a modern laptop was
systematically studied. Results indicate that wireless communication has been a significant
consumer of power, particularly obvious power hogs being the display, graphics card, and
processor. A subsequent study of energy consumption of portable devices using a remote cloud
application compared to local execution revealed some interesting facts.
vi
TABLE OF CONTENTS
Chapter
1.
INTRODUCTION ...............................................................................................................1
1.1
1.2
1.3
1.4
1.5
1.6
2.
What is Energy Efficiency? .....................................................................................1
Why Energy Efficiency? ..........................................................................................1
Current Issues ..........................................................................................................1
Problem Description and Contribution ....................................................................2
Organization of Thesis ............................................................................................3
Cloud Computing and Portable Devices ..................................................................4
MOTIVATION ...................................................................................................................6
2.1
2.2
2.3
2.4
2.5
3.
Page
Hardware and Software............................................................................................6
Procedure .................................................................................................................7
2.2.1 Hard Disk Drive ...........................................................................................7
2.2.2 LCD Display ................................................................................................9
2.2.3 Graphics Card ............................................................................................11
2.2.4 Wireless Card .............................................................................................12
2.2.5 Optical Drive ..............................................................................................12
2.2.6 I/O Ports .....................................................................................................13
2.2.7 Bluetooth ....................................................................................................15
Discussion of Results .............................................................................................16
Limitations .............................................................................................................20
Analysis..................................................................................................................20
LITERATURE SURVEY ..................................................................................................22
3.1
3.2
3.3
3.4
3.5
3.6
Introduction to Cloud Computing ..........................................................................22
Need for Cloud Computing ....................................................................................24
Related Work ........................................................................................................25
3.3.1 Efficient Wireless Communication ............................................................25
3.3.2 Power Management ..................................................................................26
3.3.3 Energy Efficiency in Clouds ......................................................................26
3.3.4 View of Power Efficiency on Mobile Applications ...................................27
3.3.5 Energy Efficiency Cloud Computing.........................................................29
3.3.6 Power Consumption in Thin Clients Due to Protocol Data
Transmission ..............................................................................................31
Local Computation and Communication ...............................................................33
Google Docs vs. Microsoft Office .........................................................................33
Office Live vs. Google Docs .................................................................................34
vii
TABLE OF CONTENTS (continued)
Chapter
4.
METHODOLOGY AND RESULTS ................................................................................36
4.1
4.2
4.3
4.4
4.5
5.
Page
Objective ...............................................................................................................36
Methodology ..........................................................................................................36
4.2.1 Hardware and Software..............................................................................36
4.2.2 Energy Measurement .................................................................................38
4.2.2.1 Windows .....................................................................................39
4.2.2.2 Ubuntu.........................................................................................41
4.2.3 Reproducibility ..........................................................................................42
Results ....................................................................................................................43
4.3.1 Ubuntu 9.....................................................................................................43
4.3.2 Windows 7 .................................................................................................44
4.3.3 Comparison of Operating Systems ............................................................44
Implications............................................................................................................45
Limitations .............................................................................................................45
CONCLUSION AND FUTURE WORK .........................................................................46
REFERENCES ..............................................................................................................................47
APPENDIX ....................................................................................................................................51
viii
LIST OF TABLES
Table
Page
1.
Description of IBM Lenovo SL400 Laptop Components....................................................6
2.
Hard Drive Power Consumed for Each Operation .............................................................9
3.
LCD Power Consumed for Each Operation.......................................................................10
4.
Wireless Card Power Consumed for Each Operation ........................................................12
5.
I/O Ports Power Consumed for Each Operation ................................................................15
ix
LIST OF FIGURES
Figure
Page
1.
Global temperatures .............................................................................................................2
2.
SATA extension cable (22 pin)............................................................................................7
3.
Hard drive setup ...................................................................................................................8
4.
Time (sec) vs. energy (mWh) for read-write operation ......................................................8
5.
LCD time (min) vs. remaining battery capacity (%) for minimum and maximum
brightness ...........................................................................................................................10
6.
LCD time (min) vs. energy (in mWh) for red, green, and blue background .....................11
7.
Graphics card: Time (min) vs. battery (%) ........................................................................11
8.
Wireless card: Time (min) vs. battery (%).........................................................................12
9.
Optical drive: Time (min) vs. battery (%) .........................................................................13
10.
I/O setup .............................................................................................................................14
11.
I/O port: Time (min) vs. battery (%) ..................................................................................14
12.
Bluetooth transmission: Distance (inches) vs. energy consumed (mWh) .........................16
13.
Bluetooth reception: Distance (inches) vs. energy consumed (mWh) ...............................16
14.
Operation: Idle (low brightness) ........................................................................................17
15.
Operation: Idle (full brightness).........................................................................................17
16.
Operation: Hard drive (copy) .............................................................................................18
17.
Operation: Optical drive (write).........................................................................................18
18.
Operation: Graphics (3D games) .......................................................................................19
19.
Operation: Wireless card(VoIP) ........................................................................................19
20.
Operation: I/O ports (iPhone charging) .............................................................................20
x
LIST OF FIGURES (continued)
Figure
Page
21.
Power consumption of mobile devices ..............................................................................28
22.
Snapshot of Google Docs...................................................................................................38
23.
Windows: Time vs. packets per sec ...................................................................................39
24.
Windows: Packet sequence vs. delay (sec) ........................................................................40
25.
Windows: Packet length vs. number of packets ................................................................40
26.
Ubuntu: Time vs. packets per sec ......................................................................................41
27.
Ubuntu: Packet sequence vs. delay (sec) ...........................................................................41
28.
Ubuntu: Packet length vs. number of packets ....................................................................42
29.
Ubuntu with cloud and non-cloud readings .......................................................................43
30.
Windows 7 with cloud and con-cloud readings .................................................................44
31.
Cloud computing: Windows vs. Ubuntu ............................................................................45
xi
CHAPTER 1
INTRODUCTION
1.1
What is Energy Efficiency?
Energy efficiency is defined as the ability to move or change the greatest amount of
matter with the least amount of energy. Sportsmen like athletes and swimmers use their bodies in
an efficient manner. Swimmers float through the water with less effort making sure they are not
splashing but simply kicking their legs. Basketball players jump for shots in an efficient way
such that the ball is moving in the same direction as their bodies so that less energy is spent
without using extra motion.
1.2
Why Energy Efficiency?
As technology increases, all electronic devices are becoming super efficient. Therefore,
when old devices are replaced with newer appliances, special care should be taken to purchase
those that have the ENERGY STAR label. Using energy-efficient devices not only saves in
energy costs but also reduces global warming and the earth’s burden on energy resources [1].
1.3
Current Issues
Global warming is a major concern these days. There has been a substantial increase in
surface-temperature patterns since the mid-twentieth century [2]. The temperature increase over
this period has been almost 0.74 ± 0.18ºC. Greenhouse gas emissions have increased because of
human activity and the burning of fossil fuels.
Gartner Inc. claims that there has already been a use of 1 billion personal computers
(PCs) worldwide, and at an annual growth of 12%, this number will likely reach two billion by
early 2014. The United States, Europe, and Japan together account for 58% of the installed PCs
worldwide. Generations continue to change, the latest being called the M-Generation who
1
multitask using new gadgets. A company gain has become others’ burdens as each television
company spends millions on advertising to display products at outlets and at the same time
campaigning for the use of ENERGY STAR devices. Eco-friendly devices have been on the
main agenda for most key market players. If not taken to heart, the historical increase of global
temperatures, as shown in Figure 1, will definitely adversely affect the world in the years to
come.
Figure 1: Global temperatures [2]
1.4
Problem Description and Contribution
As discussed earlier, energy efficiency has been a key factor in many issues. Due to the
boom in information technology (IT) over the past decade, the use of portable devices has
increased globally [3]. Millions of PCs worldwide that are left in an idle state all night could be
2
moved to an off-state mode. Small companies are now developing solutions to power off PCs
from a remote location, which could save millions of dollars in budgets annually.
The share of overall global consumption by portable devices, such as laptops and mobile
devices, is approximately 1%. Statistics show that desktop sales have been surpassed by laptop
sales, due to current extensive research on mobile computing devices. Portable devices have
increased exponentially over the years to 8.5 billion, eight times more than the last decade. The
life of the laptop battery has been a major concern as technology and applications are added to
portable devices. This issue can be viewed in two primary ways: either extending the life of
batteries or reducing those areas that have greater carbon emissions.
Recent carbon emissions from standby devices alone can be measured in tons, energy that
could have provided power and saved considerable resources in poor countries for years. In
addition, organizations are migrating to Web 3.0 and web-based applications, which use a
tremendous amount of computing power. Clearly, growth in this economic sector could occur if
energy is saved through cloud applications, that is, Web-based processing whereby shared
resources, software, and information are provided to computers and other devices on demand. In
this research, one of the main focuses is the study of energy patterns of portable devices such as
laptops on office productivity tools and whether remote execution is more energy efficient than
local execution.
1.5
Organization of Thesis
The initial emphasis of this thesis is on the driving factors that led to this research,
followed by the reason to choose cloud computing and discussion on some closely related
research papers. Then the thesis focuses on how experiments were performed, and the results of
3
cloud and non-cloud scenarios are compared. Finally, discussion focuses on the conclusion of the
research with respect to energy expenditure in terms of various scenarios.
1.6
Cloud Computing and Portable Devices
Cloud computing is used as a metaphor for the Internet, where resources, information,
software, and hardware are shared. The main advantages of using cloud computing are
scalability, instantaneousness, and lower maintenance costs. Today, cloud computing has been
employed by many major companies, such as Google, IBM, Microsoft, etc. Cloud computing is
used on a daily basis; however, it may not be noticeable at all [4].
Imagine that you have been on a trip and have an album of pictures saved on your local
computer. Now you want to share them with friends. Due to the many photo-sharing websites,
like Flickr and Picasa, it is easy to upload the pictures online and then share the link of the
album’s location. This is cloud computing, where the album is stored in a data center online in a
cloud and thus is instantaneous. Also, imagine that at work you have an impending deadline and
want to prepare documentation. At the same time, you want to share this documentation with
someone on your team in order to proof it. Today, it is easy to do this with cloud applications
like Google Docs, Office Live, etc.
Cloud computing is typically a client-server architecture, where the client can be any
portable device, such as a laptop, phone, browser, or any other operating system (OS)-enabled
device. The components involved in the client-server models are application, platform, and
infrastructure. There is no need for installing and running the applications, buying the hardware
or software platforms, or even the data center or server usage, which is billed based on
utilization.
4
These days many portable devices exist where users want to share documents, email, and
surf the Internet instantaneously without planning. However, the problems with such devices as
the iPhone, iPad, and netbooks are their lack of data storage capacity. They do not have jumpdrive slots to copy the data when storage is totally utilized, and information must be regularly
backed up. With these devices, data needs to be saved instantaneously to a safe and trusted
location like a cloud or a desktop computer system in a home. Tests must be performed to see if
any energy is saved with remote computation as compared to local computation.
5
CHAPTER 2
MOTIVATION
Statistics from past years show that desktop sales have been surpassed by laptop sales
worldwide. Laptops have been broken down into components and the energy expended by each
component has been measured. The results for each component have been studied, and these
results provide useful information as background for the experiments in this study.
This chapter first introduces the hardware and software used in this work. Then, the
procedure is explained in detail, and graphs of results obtained for each component are
discussed. Finally, limitations and the breakdown of components are explained.
2.1
Hardware and Software
The laptop used in the work was an IBM Lenovo SL400 [5] with a 14-inch display. The
operating system loaded on it was Ubuntu 9.04 [6]. The different hardware components and
details are shown in Table 1.
TABLE 1
DESCRIPTION OF IBM LENOVO SL400 LAPTOP COMPONENTS
Component
Processor
Memory
Hard Drive
Optical Drive
Wireless Networking
Display
Battery
Ports
Graphics
Details
Intel Core 2 Duo 2.4 GHz
2 GB
160 GB
DVD + RW
Intel 4965AGN (802.11 a/b/g/n Wi-Fi) Bluetooth 2.0
14.1‖ 1280 x 800
6-cell Li-Ion
4 USB 2.0 ports and more
NVIDIA GeForce 9300M GS 256 MB
®
™
The laptop’s display technology was WXGA TFT active matrix [7], which supports 24bit color (16.7 million colors). The setup used for conducting the experiments was a multimeter,
6
to obtain the direct readings, and the ―
state‖ of the battery at different stages of the experiment,
for the rest. The stress to any system is graphics, which was achieved using 3D games. Direct
measurements were done for the hard drive and the input/output (I/O) ports.
2.2
Procedure
2.2.1 Hard Disk Drive
The hard disk drive connected to the laptop under examination was a 7,200 rpm drive
with 160 GB capacity. In its original configuration, the hard disk drive was connected to the
laptop through a 22-pin SATA connector. In order to measure the current and voltage across the
hard-drive terminals, a 22-pin SATA extension cable was used, thus facilitating the use of a
multimeter for measuring the current, as shown in Figure 2.
Figure 2: SATA extension cable (22 pin)
At first, the power consumption of the hard drive was studied by running the standard
read, write, and copy operations on the disk and checking the power drained from the battery to
estimate the effect of hard drive utilization on the battery. A 697.9 MB Ubuntu International
Organization for Standardization (ISO) image file was used for performing the tests. The test bed
is shown in Figure 3.
7
Figure 3: Hard drive setup
As shown in Figure 4, the read-write operation with maximum brightness consumes more
energy. Table 2 contains the readings of direct measurements of different operations. The results
idle‖
obtained were highest for the ―
copy‖ operation, which was 3.37 W, and lowest for the ―
operation, which was 2.13 W.
40670
Read-Write (Full Brightness)
40660
Energy (in mWh)
Read-Write (Low Brightness)
40650
40640
40630
40620
40610
40600
0
1
2
3
4
5
6
7
8
9
10
11
Time (min)
Figure 4: Time (sec) vs. energy (mWh) for read-write operation
8
TABLE 2
HARD DRIVE POWER CONSUMED FOR EACH OPERATION
Operation Power Consumed (W)
Idle
2.13
Read
2.73
Write
3.32
Copy
3.37
2.2.2 LCD Display
The energy consumed by the laptop was measured using an in-built procedure taken from
the advanced configuration and power interface (ACPI) of the Ubuntu operating system.
Measurements were done using a different approach. The display was isolated by removing the
hard drive and optical drive, and turning off the network and Bluetooth devices. The laptop was
booted using a Universal Serial Bus (USB) drive. Once the laptop was up and running, the USB
drive was removed, leaving the laptop with only the liquid crystal display (LCD), central
processing unit (CPU), and random-access memory (RAM). Here CPU and RAM took a
minimal amount of energy since nothing was processed.
Brightness was kept at a maximum level, and appropriate readings were taken from the
ACPI procedure every minute for more than an hour. The experiment was repeated for
brightness with minimum level, using a black and white background. The LCD power
consumption is provided in Table 3. Observations show that the power consumed was 15% more
when the brightness level was at maximum rather than minimum.
9
TABLE 3
LCD POWER CONSUMED FOR EACH OPERATION
Operation
Power Consumed (W/min)
Black Background
White Background
Default Background (Full Brightness)
Default Background (Low Brightness)
0.210
0.202
0.207
0.180
It can also be seen that a black background consumed more power than a white
background. This could have been due to the twisted nematic field effect in the LCD display [8].
Therefore, in this laptop, a black background took 3.96% more energy than a white background.
Figure 5 shows the reduction in battery capacity with maximum and minimum brightness levels.
100
Remaining_Battery_Max_Brightness
Battery capacity (%)
95
Remaining_Battery_Min_Brightness
90
85
80
75
70
65
60
55
0
5
10
15
20
25
30
35
40
45
50
55
60
65
Time (minutes)
Figure 5: LCD time (min) vs. remaining battery capacity (%)
for minimum and maximum brightness
Individual backgrounds such as red, blue, and green were also tested. Observations show
that energy consumption was the same in all three cases, as shown in Figure 6, where the graphs
for all three colors overlap.
10
52000
Red
Energy (in mWh)
51500
Blue
Green
51000
50500
50000
49500
49000
0
1
2
3
4
5
6
7
8
9
10
11
12
Time (minutes)
Figure 6: LCD time (min) vs. energy (in mWh)
for red, green, and blue background
2.2.3 Graphics Card
A NVIDIA graphics card was used in this experiment. Three-dimensional games were
played using a Wi-Fi and later with the Wi-Fi turned off. The 3D game used without Wi-Fi was
Warezone 2100 [9] and the game with Wi-Fi was Evony [10]. Additionally, a 2D game, such as
Solitare [12], was used for comparison. Figure 7 explains more about the comparison of the three
stages of stress on graphics. As can be seen, the 3D game with Wi-Fi consumed more energy
than any other methods.
Battery capacity ( %)
100
95
90
85
3D game with Wi-Fi
3D game without Wi-Fi
2D Game
80
75
0
5
10
15
20
25
Time (minutes)
Figure 7: Graphics card: Time (min) vs. battery (%)
11
30
35
2.2.4 Wireless Card
The wireless local area network (LAN) standard that was considered for testing the
laptop was IEEE 802.11g [12], for which the standard transmission rate is 54 Mbps. The
applications tested were file transfer profile (FTP), and voice and video streaming. To check the
power consumption of the FTP application, a 690-MB file was downloaded. The voice
application was tested using Skype [13]. The power consumed by the laptop when each
application was run is shown in Table 4, and battery consumption is shown in Figure 8.
TABLE 4
WIRELESS CARD POWER CONSUMED FOR EACH OPERATION
Application
Power Consumed (W)
Wireless-Off
17.3
FTP
18.732
Video Streaming
20.681
VoIP Application
20.629
Battery capacity (%)
100%
90%
80%
70%
60%
Wireless card off
50%
Wireless card on
40%
FTP download
30%
Voice-Chat
20%
0
10
20
30
40
50
60
70
80
90
100
110
120
130
Time (minutes)
Figure 8: Wireless card: Time (min) vs. battery (%)
2.2.5 Optical Drive
The component used to test the optical drive was a DVD + RW drive, which is capable of
writing and reading from both 8x as well as 16x disks. This component was part of the Lenovo
SL400 laptop. It used a SATA cable with embedded power cords [14] for transmitting data.
12
Testing was performed for two kinds of operations on this device: one for the read
operation, the other for the write operation. During the initial part of the setup, testing was done
on a small amount of data (600 MB) at different brightness levels of the LCD display. One key
observation here was that low brightness not only consumed less energy, but less time was used
to perform the operation. Results from both read and write operations are shown in Figure 9.
100
Battery capacity (%)
99
DVD Read
DVD Write
98
97
96
95
94
93
0
1
2
3
4
5
6
7
8
9
10
Time (minutes)
Figure 9: Optical drive: Time (min) vs. battery (%)
2.2.6 I/O Ports
In the initial setup, energy consumption measurements on the USB port [15] were taken
by playing a movie continuously for an hour using a flash drive. This setup is shown in Figure
10. Tests were conducted by setting the brightness level to maximum with the Wi-Fi switched
off. The remaining charge on the battery was measured every five minutes, and readings were
tabulated. The graph of this is plotted in Figure 11, which shows a linear type of decrease, i.e.,
the charge on the battery decreases gradually as the movie was played continuously.
13
Figure 10: I/O setup
100.00
95.00
Battery capacity (%)
90.00
85.00
80.00
75.00
70.00
65.00
60.00
55.00
50.00
0
5
10
15
20
25
30
35
40
45
50
55
60
65
Time (minutes)
Figure 11: I/O port: Time (min) vs. battery (%)
Some USB devices can shorten battery life dramatically when used for longer periods of
time. A set of standard USB devices seen in daily life was used to test this. A USB extension
cable connected and measured the power consumption of USB devices with the aid of an
ammeter. Table 5 shows the devices measured with different operations.
14
TABLE 5
I/O PORTS POWER CONSUMED FOR EACH OPERATION
Device
Power Consumed (W)
iPhone
2.475
4 GB Flash Drive (idle state)
0.297
4 GB Flash Drive (data transfer)
0.445
4 GB Flash Drive (watching movie)
0.297
External HDD (with ext power source)
0.098
Portable HDD (idle state)
1.881
Mouse (idle state)
0. 148
Mouse (working mode)
0.270
2.2.7 Bluetooth
To measure the energy consumption of a Bluetooth device, a BlueSoleil dongle [16]
version 2.0 was used for transferring files to and from the laptop and mobile devices. The mobile
devices used for testing were a Sony Ericsson W580i (Bluetooth version 2.0), MotoV3
(Bluetooth version 1.1). Due to the low transmission rate of the Bluetooth version 1.1, the time
taken to transfer data was greater, which lead to more energy consumption. For version 2.0, the
energy consumed was comparatively less due to the faster transmission rate. Readings were
taken with full brightness and with the Wi-Fi turned off. A comparison of transmission and
reception for both versions is plotted in Figures 12 and 13, respectively. It is evident that
receiving data consumes more energy than transferring data. For Bluetooth version 1.1, the
reception was very high compared to that of version 2.0.
15
550
Transmission
530
Energy (in mWh)
510
490
470
Energy Consumed with v2.0
450
Energy Consumed with v1.1
430
410
390
370
350
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
Distance (in)
Energy (in mWh)
Figure 12: Bluetooth transmission: Distance (in) vs. energy consumed (mWh)
690
670
650
630
610
590
570
550
530
510
490
470
450
Reception
Energy Consumed with v2.0
Energy Consumed with v1.1
0
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
Distance (in)
Figure 13: Bluetooth reception: Distance (in) vs. energy consumed (mWh)
2.3
Discussion of Results
Figure 14 shows the breakdown of components in an idle system with low brightness and
also the total system power observed during ―
rest.‖ As can be seen, the hard drive consumed
much more energy than the other components.
16
IDLE (Low Brightness) 12.04 W
15%
LCD
HDD
Optical
18%
Rest
63%
4%
Figure 14: Operation: Idle (low brightness); system power 12.04 W
Figure 15 shows the breakdown in components for the experiment conducted for an idle
system with full brightness. Here, the LCD dominated in the breakdown of energy.
Idle (Full Brightness) 14.88 W
LCD
31%
HDD
Optical
Rest
51%
15%
3%
Figure 15: Operation: Idle (full brightness); system power 14.88 W
Then experiments were conducted on the hard drive when the ―
copy‖ operation was
performed. This experiment was performed at full brightness. Figure 16 shows that the LCD also
used a considerable amount of energy.
17
Hard Disk (Copy) 15.37 W
31%
LCD
44%
HDD
Optical
Rest
22%
3%
Figure 16: Operation: Hard drive (copy); system power 15.37 W
The optical drive took less energy when performing a read operation since it buffered
data into the RAM when it was inserted in the system. A significant amount of energy was
consumed when a write operation was performed, because the disk needs to be spinning
continuously. Results shown in Figure 17 indicate an increase in energy consumption.
Optical Drive (Write) 20.88 W
22%
37%
LCD
HDD
12%
Optical
Rest
29%
Figure 17: Operation: Optical drive (write); system power: 20.88W
Graphics play a significant role in the amount of energy consumed on a system when 3D
acceleration is used. Under idle conditions it will be around 5% of the overall consumption in a
laptop. When it is stressed over a period of time, as the result of using 3D games, the rate at
18
which the battery drains will be higher. This can be observed in Figure 18, which shows the
energy consumed when 3D games were played over a period of time. Graphics dominate the
chart significantly.
3D Graphics 23.48 W
8%
12%
LCD
2%
HDD
Optical
Graphics & Rest
78%
Figure 18: Operation: Graphics (3D games); system power: 23.48 W
A wireless network interface card (NIC) card was tested with different types of
applications like FTP, voice over Internet protocol (VoIP), and video streaming. Of all the tested
applications, VoIP over a period of time consumed more energy, as shown in Figure 19.
Wireless (VOIP) 19W
25%
LCD
40%
HDD
Optical
11%
Voip
Rest
2%
22%
Figure 19: Operation: Wireless card (VoIP); system power: 19 W
Various experiments were conducted for the I/O ports by transferring data from a USB,
watching a movie from it, and charging a mobile. Today, mobile units are sold with a mobile
19
charging option through a USB by simply connecting it to a laptop, but a significant amount of
energy is used in this process. Figure 20 shows the results of this experiment.
USB (iPhone charging) 17.355 W
27%
LCD
HDD
43%
Optical
USB
13%
14%
Rest
3%
Figure 20: Operation: I/O ports (iPhone charging); system power 17.355 W
2.4
Limitations
The accuracy of these experiments was limited by various factors. When measurements
were made via the multimeter, at times there was a marginal error of +/- 3mA. Also, some results
were obtained using the subtractive approach, so the results obtained could vary with an error
rate of less than 1% [17]. Finally, the results were obtained for some components, and the
integration of other components was termed as ―
rest‖ in the pie charts, which means the rest of
the system.
2.5
Analysis
The main purpose of this work was to obtain the component-wise energy breakdown of a
laptop. Energy patterns were obtained for the following components: hard disk, display, graphics
card, wireless card, optical drive, I/O ports, and Bluetooth. Due to some limitations,
measurements were performed using the direct approach and subtractive measurement approach.
Thus, the results obtained suggest that the display, graphics card, and processor consume a
20
significant amount of energy in a system. Surprisingly, the wireless card was found to consume
more energy, based on various applications tested. This amount of energy consumption is next to
that of the display, graphics card, and processor.
In a system, the processor, display, and graphics card components vary depend on use. If
the system selected is a server for the data center, then the processing capabilities change for that
amount of load handled for each process. If the system is a 3D designing platform, then the
graphics used are intensive on the laptop. If the system is a gaming hub, then the display
changes quite often with a graphics card, and the display uses a major amount of energy.
The wireless card is almost the same on all laptops for all applications. Operating systems
that use cloud applications are released by major companies because there is a future for them.
For these applications to operate online, a wireless card is the key component in the laptop. Also,
the results obtained from this chapter and the non-variable characteristics shown by a wireless
card on the laptop helped to move this study forward relative to exploring future energy
concerns. One such concern identified was local computation versus cloud computation for
similar tasks and is discussed in detail in later chapters.
21
CHAPTER 3
LITERATURE SURVEY
3.1
Introduction to Cloud Computing
Cloud computing, or simply cloud, is an economic and business model. The answer to the
common question of whether data centers can be replaced by cloud computing is partially yes
and partially no [18]. Cloud computing is the future of the Internet, which changes everything
else. A computing infrastructure has already been designed, applications have been developed,
and both business and personal collaboration have been delivered over the cloud as a service.
Cloud computing is offered in different forms:

Public

Private

Hybrid (combination of public and private)
Cloud computing can be compared to a fluid, which expands and contracts, i.e., it is
elastic. This is the main reason for users to ―
move into the cloud.‖ Storage, network,
infrastructure, users, and applications can be added and subtracted, and it is easy to do from
personal desktop computing, since these services are fast and dynamic companies are moving
forward. Not all data centers and their computing assets can be moved over the cloud, and a
close examination is required depending on the business deployed.
Cloud computing can offer many services over the Internet such as networking, storage,
delivery of software, infrastructure, application programming interfaces (APIs), etc. Basically,
cloud computing consists of three types of participants:
1. Service provider, who is responsible for IT asset management and maintenance.
2. Management, who will be ensuring service and security for the data.
22
3. End user, who is not concerned about the underlying infrastructure or network but rather
the performance of the service that is utilized.
Many new services are deployed over the cloud. In the past, stocks were traded at
financial market centers, but today, they can be executed on any personal PC without worrying
about security. The demand for social networks, such as Facebook and LinkedIn, is increasing
drastically. Companies have started using the cloud to share data and collaborate using personal
applications, examples of which are Google Docs and Office Live. For personal accounts, these
services are accessible free of cost. Billing and metering as a per service usage, pay-as-you-go
model, are offered for these types of services. Many APIs have been developed to attract users
and businesses. Also, users are able to instantaneously share personal photos and music over the
cloud, such as with Picassa and Flickr. For these reasons, cloud computing is scaling up
elastically in order to provide these many services.
Cloud offerings can be broadly categorized into three basic services:
1. Software as a Service (SaaS), which offers multiple office applications depending on the
type of business. The virtual private network (VPN), now accessed on the public
network, was used previously because of its security and reliability.
2. Platform as a Service (PaaS), which offers development platforms for IT users to create
or develop business applications. The Google API engine and Salesforce.com are
examples of this service having a full development environment.
3. Infrastructure as a service (IaaS), which offers computing resources and storage
capabilities that IT and other developers use for custom business solutions.
Cloud services are similar to cable TV, which offers certain channels depending on the
package selected. Watching a movie on demand and paying per use is similar to the way the
23
cloud functions. For these types of services, security is a concern, and well-defined planning of
the following assures companies of a secure network:

Access control, for protecting the security of resources.

Identity management, any hardware component, or application service authorization.

Authorization and authentication, where only suitable people are able to modify or create
data and applications.
3.2
Need for Cloud Computing
When starting a new organization, the traditional computing services involved are
implementing hardware, installing software, planning disaster recovery, designing networking,
and providing uninterrupted uptime even during scheduled maintenance. Almost 42% of an
organization’s budget is spent annually on these services, and the remaining percentage is spent
on heating, air conditioning, taxes, and labor costs. Labor costs alone account for 40% of the
budget. The same factors observed for cloud computing services include the following: 48% for
computing costs; 20% for cooling and power, since most service is provided over the cloud; and
around 6% for labor costs, since the person-to-server ratio improves almost ten times for cloud
services.
The portion of an IT budget spent annually to run and maintain a cloud system is 70% to
80%, and to build and implement the new capabilities is only 20% to 30%. By using cloud
services, maintenance costs are reduced, thus focusing on implementing new capabilities.
Location also plays an important role in designing a cloud system. Land and building costs are
more important and should not be taking more of the annual budget.
An entire data center that is operated by a cloud or locally computed can be destroyed if
security is not managed well. Applications should be written for PaaS and SaaS when firewalls
24
are designed to operate well. These are some of the factors that major companies such as Google,
Microsoft, IBM, Amazon, etc., are looking for these days. The migration to cloud services has
begun.
1.3
Related Work
Most recently, considerable applications and software have been designed to automate
most of them. Major changes are primarily occurring because of the various research areas in the
wireless field. Since more and more technology is being added, the time required for each device
to last without a power outlet is becoming limited.
3.3.1 Efficient Wireless Communication
Some researchers [19] contend that wireless communication is a power-hungry
application, the key reason being that it radiates power in all directions. For this kind of scenario,
they developed a proposal to use an omni-directional beam switch with a multiantenna system
for efficient directional communication. They even designed and implemented a prototype on
Rice University’s Wireless Open Access Research Platform (WARP) to show its performance. It
was tested on a VLC Media Player, using the data transfer of a large file and also a remote
desktop unit for indoor and outdoor use. Indoor refers to a cross-office scenario and inside a
hallway of a building, whereas outdoor refers to a university campus. The results were
interesting: an almost 20% savings of the total energy consumption under extreme conditions for
more mobility and total transmit power. In this technique, more focus was placed on the antenna
part of the wireless card and never involves a cloud services scenario for more propagation when
remote computation is done.
25
3.3.2 Power Management
An entirely different paper that focused on the power-saving mode aided in the
understanding of how much time is spent by a wireless device in the idle state while still
consuming considerable energy to maintain it. Normal medium access control (MAC) protocols
have a power saving mode (PSM), which is effective for elongated idle periods such as >100 ms.
Measurements show that almost 80% of the time was spent on waiting for the frames with idle
intervals less than 200 ms. In this research [20], an implementation was done to demonstrate a
busy time using micro power management. Simulations observed in NS2 and traces were taken
using a ―W
ireshark‖ application, which reduces energy consumption by almost 30% without
degrading quality. These two studies mainly focused on saving energy, which was advantageous
but not applicable to cloud computing services, because the idle time was less when running
these kinds of services on cloud-based operating systems since communicating occurs often.
3.3.3 Energy Efficiency in Clouds
Servers on the clouds can be both physical and virtual machines. For large clouds the
major concern is power consumption. Today, computing devices are able to operate at different
frequencies. Abdelsalam et al. [21] found a mathematical relationship between the number of
servers operated and the service level agreements (SLAs). Along with these various services
such as hardware failures, change in management and devices were focused upon.
Cost-efficient information services and IT management should be able to deliver over a
cloud system. Many companies like Google, IBM, Amazon, Yahoo, etc., are available over the
cloud. The major challenges are the IT computing environment and power management for these
companies. High power consumption is not only limited to energy costs; the cost of the initial
investment, such as a cooling system, should also be taken into consideration. All of these
26
operational costs could be reduced using the SLA technique. Several hardware and software
techniques can be used for decreasing operational costs:

Hardware techniques: dynamic frequency scaling, processor throttling and low-power
DRAM states.

Software techniques: operating system virtualization, advanced configuration and power
interface (APCI), and standards.
System administrators have a mathematical model to compute the frequencies of the
servers that should be used. Updating the utilized hardware and software is referred to as change
management. Servers should be running at top speed (frequency) so that an application will be
completed as soon as possible. The reason behind this approach was to allow dedicated servers to
be idle for longer periods, thus improving the overall energy consumption. The authors of this
paper decided to relax some of the model assumptions that they previously made. This paper was
able to focus on energy over the clouds but made several assumptions.
3.3.4. View of Power Efficiency on Mobile Applications
The evolution of both portable and mobile communication systems has led to an increase
in hardware and software demands, together increasing significant market penetration. Power
efficiency metrics have been applied on different wireless systems, and experimental data
obtained in the monitoring application has been analyzed. The scenarios used for various
communication systems involve the wireless local area network (WLAN), Bluetooth, global
system for mobile communication (GSM), etc.
Today, more utilities are embedded in mobile devices such as WLAN, multimedia,
Bluetooth, GPS, VoIP, etc. Statistics have already shown that there has been a growth in value of
these mobile devices: about eight times in five years (2005–2010), i.e., $1 billion to $8.5 billion.
27
The number of new applications that runs on mobile phones has increased dramatically. Battery
capacity has been a major problem for these devices. Reports show that IT consumes almost 2%
of energy, which could be controlled by reducing the processing core frequency by about 20%
and adding new core processors. Various consumption circuits, like the processor, memory,
communication, and display, show how much energy is used by each component. Different
communication patterns are observed for various communications such as transmission control
protocol/internet protocol (TCP/IP), GSM, Wi-Fi, and Bluetooth. The pie graph shown in Figure
21 is taken from Marcu et al. [22] and shows the power consumption of mobile devices
Figure 21: Power consumption of mobile devices [22]
Their main objective was to design and implement a software framework for poweraware mobile applications and to increase energy efficiency. On a daily basis, around 99% are
under GSM and 49% are under Wi-Fi. Based on the type of application usage, such as email,
feeds, multimedia, web, browsing, data transfer, VoIP is involved, the widespread use of these
applications has led to increased energy consumption.
The metrics used for the patterns were battery discharge rate, current consumption,
remaining battery, capacity, etc. The various software utilized for the prototype were C++ and
Visual Studio 2005, with testing done on Windows 32 (Win32), Windows Mobile 5.0, etc. The
28
framework designed has a battery monitor, CPU monitor, CPU parameters, wireless monitor, and
workload generator (different traffic patterns). Simply adding Bluetooth can add up to 13%
energy consumption in the same amount of time, almost 50% being utilized for WLAN as
compared to the idle power consumption. Security is a major concern in cloud computing;
therefore, added encryption increases the energy consumption. This paper concludes that there is
no meaningful drop in default power management techniques on the WLAN for application-level
communication. To have an impact and thus a better use of energy, it is necessary to involve
both system and application.
This paper focused mainly on energy patterns for mobile devices, which is helpful in
identifying the facts and comparing the results obtained with the latest hardware, but it did not
suggest a significant reason for doing this.
3.3.5 Energy Efficiency Cloud Computing
Many information and communication technology (ICT) systems lead to an increase in
energy costs and greenhouse gas emissions. Almost 53% of Amazon’s budget accounts for the
cost and operation of servers. Around 42% are energy-related costs: direct power consumption
(19%) and infrastructure cooling (23%). A data center energy forecast report suggests that almost
20% of energy savings can be achieved at the server and network, and around 30% can be
achieved at the cooling systems.
A positive impact is possible if a fraction of the energy savings can be achieved on
carbon and financial savings. Other places where energy can be saved are by using energyefficient hardware during data replication and task transference.
In the coming years,
virtualization will play a key role since less hardware will be involved. Different forms of
virtualization are full virtualization, hosted virtualization, and operating system layer
29
virtualization. Many energy-hungry components like discs, memory, and network devices
depend on CPU utilization. Idle energy also uses up to 60% of CPU utilization; in these types of
situations, jobs should be dispatched to a small set of active servers while others are down to
lower state for economic criteria and energy as criteria.
Much energy is used up in wired and wireless networks, where minimization is required.
Annual consumption by the Internet alone is 860 TWh, and ICT estimates that around 3 TWh
energy was consumed by the Vodafone Group radio access in 2006. A new tradeoff arises when
high transmission power is utilized and the number of hops is reduced. Conversely, the greater
the number of hops leads to greater energy consumption. Infrastructure networks are energy
consumptive because of storage, processing elements, job scheduling, virtualization, and
migration. Some areas where energy can be utilized are:

Energy efficiency can be done in various places.

A study of individual components while designing hardware and software is essential.

Unused servers can be turned off to save considerable energy.

Energy can be saved in networks and protocols.

The effect of Internet applications, digital cameras embedded in mobiles to
environmental sensors to Web 2.0, end users are generating and interconnecting over the
cloud
Almost 85% of Internet traffic is dominated by peer-to-peer and video-on-demand
services. If cloud computing becomes a platform, then there would be significant increase in data
flow for tasks like producing and accessing information, and sharing data.
In conclusion to this paper [23], the authors suggest cloud computing along with
virtualization as a way to move forward:
30

Significant tradeoffs between performance, QoS, and energy efficiency should be
maintained, and the main sources of energy consumption should be identified.

Insight on the energy savings that can be achieved on large-scale computer services that
integrate communication needs to be offered.
By taking these stems, a significant impact will be seen on the following:

Reduction in hardware- and software-related costs.

Reduction in energy consumption due to communications.

Improvement in load balancing, and hence QoS and performance of single or federal data
centers.

Reduction in carbon dioxide and greenhouse gas emissions resulting from data centers
and network.
Improvements can be done by reducing energy utilization for transportation and working
on smart solutions for e-work, e-learning, and smart climate homes. This paper recommends
moving towards virtualization and cloud computing, but no practical implementation or
prototypes is provided.
3.3.6 Power Consumption in Thin Clients Due to Protocol Data Transmission
Applications are run at the user terminal, but in this particular setup they are executed at
the network server so that the processing time spent would be reduced and tested on thin clients
since they are energy efficient. The concept proposed is advantageous for mobile users because
only the basic functionality and processing power is required at the user terminal, and redundant
hardware can be removed, thus resulting in thin lightweight devices. Since all data and logic is
shifted to the network, the user terminal should be connected to the network at all times. All the
mobile users will have connectivity to the application server over a wireless interface such as
31
UMTS, Wi-Fi, etc. Various scenarios are tested such as text editing, browsing on websites with
static text, and running animations. The influence of the multiple wireless channel parameters is
investigated, and a previously developed cross-layer approach situated at the MAC layer and the
physical (PHY) layer is compared in terms of power.
In a mobile terminal scenario [24], the largest consumer of energy was the wireless
platform, when considering application and visualizing components. To watch a video for a
minute, wireless communication takes approximately 37.7% of the total consumption.
Fortunately, a high level of scalability is offered by wireless platforms so that configurations can
be adapted to instantaneous requirements and external conditions. By exploiting all of this,
scalability through a cross-layer approach between the MAC and PHY layers can be designed.
Earlier demonstrations have shown that terminal power consumption is up to 70%, compared to
the state-of-art 802.11 link adaption at the MAC layer. Both link adaption and a cross-layer
strategy are used.
Related work by the authors showed that thin clients took 85% less power compared to
the PC (server powering, cooling). Also, some authors have suggested a cross-layer design to
improve service quality of centralized-server gaming architecture. All experiments were done in
a network simulator (NS2), and a VNC thin-client protocol architecture was used. An automatic
Xnee tool,‖ was used for each scenario. The scenarios are as follows:
replay tool, called ―

Text editing: open office used to insert the title and simulation time of 200s, insert
images, and others.

Static browsing: going to static websites and simulation time of 140s.

Dynamic browsing: sites with JavaScript, flash, and video-on-demand (VOD) for a
simulation time of 120s.
32
By using this approach, the authors were able to save 14% of the total energy
consumption. Some suggestions put forward were having dynamic sleep time schedules for
wireless that may reduce some idle times for energy savings. Having a good monitoring
framework could be helpful in reducing some energy consumption. This paper is closely related
to the work in this thesis, but it focuses mainly on the server end of the network, not on the client
side to show how energy patterns are varying.
3.4
Local Computation and Communication
Some key factors must be considered when dealing with local computation versus
communication. Hardware, software, and maintenance are the initial components of building a
network. Instead of having all computing resources on the client side, it is better to have them on
the server side in a client-server architecture. Since the server end has unlimited power and even
provides a backup for providing continuous energy, it is now required to think at the client end
where the scope of the power outlet is minimal at all times.
The next chapter deals with one such attempt to identify energy savings when
computation is done at the server end rather than the client end. Here, communication between
the two was considerably less. There are many ways to test which scenario is better, but in this
case, office productivity tools were chosen: Google Docs, Office Live, Microsoft Office, and
Open Office.
3.5
Google Docs vs. Microsoft Office
Cloud computing helps us with using normal services over the cloud. Google Docs is one
tool for doing this and has gained popularity in recent times due to its far-reaching capabilities.
When opening or editing a normal Word document sent by another person, it is necessary to
have that software running on the personal computer. If it is to be presented in an office, then it
33
is necessary to check if the software is installed and if the latest version is running in order to
view it.
With cloud computing, the document can be accessed without running the software, and
even multiple users can edit or view the document at the same time. Whereas normal documents
are saved in a personal computer, in the case of cloud computing, the document is saved on a
web server where it is accessible by everyone. In addition, different types of permission can be
specified for different users. And, since this process is operating on the cloud, there is no need to
copy files from one PC to another in order to make modifications. This process occurs in real
time and is very fast.
3.6
Office Live vs. Google Docs
Both Office Live and Google Docs are cloud-based applications, both of which are being
used often these days [25]. Although they perform similar functions, they differ in some features.
In Office Live, files cannot be modified within the live office workspace, but by selecting the
―
edit‖ option, the files can be modified by opening them in Microsoft Word.
Both applications are free services offered on the web, and both are good for cloud
computing. Some of their unique features are discussed below:

Relative to storage space, Office Live has limited space, about 500 MB, which comprises
almost 1,000 Microsoft documents, whereas Google Docs limits each user to 5,000
documents and presentations, 5,000 images, and 1,000 spreadsheets, although an actual
space is not specified.

Office Live supports Microsoft documents along with PDF files, whereas Google Docs
not only supports the above file types but also allows most open-source file types for
Word and spreadsheets.
34

Google Docs has file sizes of 500 K each, and has various file sizes for images and
presentations of 2 MB and 10 MB, respectively. In the case of spreadsheets, this
translates to a limit of 256 columns or 40 sheets, whichever is reached first. For Office
Live, the workspace for each file is 25 MB, regardless of type.

File sharing is similar in both applications, but real-time collaboration is available in
Google Docs, whereas it is not in Office Live.

For uploading multiple files, Google Docs uses Docsyncer, a third-party tool that uploads
all files from the PC to the cloud. All documents and presentations can be emailed but not
spreadsheets. Office Live gives an option to choose single or multiple documents when
adding a document if opened in Internet Explorer; otherwise, this option is not available.

Various versions are supported in both applications, since both keep track of earlier
versions and can roll back anytime.

Direct URLs, which can be bookmarked, are provided by both applications. Saving files
and exporting them in Office Live assumes Microsoft Office by default, except for PDF
files, whereas Google Docs saves or exports files as Microsoft Office, HTML, Open
Office, PDF, or text files.
35
CHAPTER 4
METHODOLOGY AND RESULTS
4.1
Objective
The objective of this research study was to run office applications remotely as well as
locally, and determine whether remote execution is more energy efficient than local execution.
4.2
Methodology
The study for office productivity tools is done by experimental approach in two different
versions, i.e., Ubuntu and Windows.
4.2.1 Hardware and Software
The hardware and software discussed in Chapter 2 were used to maintain continuity for
all experiments in this work.
Windows 7 is a new version of Microsoft [26]. Within a short span, it has gained
popularity because of its unique features; even the boot capabilities have been improved,
compared to Windows Vista. The main goal is to be compatible with hardware that supports
Vista and also provide performance improvements. This software is written in C, C++.
Ubuntu is an open-source software based on a GNU/Linux (Debian) distribution [6]. This
software is easy to install, and almost 50% of Linux desktop clients use it. There is an uptrend
because of the usage as a Web Server these days. Version 9 was used for testing purposes, and
the latest 10.04 stable release is currently out on the market. This version has a basic word
processor and networking ports that offer selection. Ubuntu supports a virtual machine program,
Wine, which is designed to run Windows OS applications. Recently, Ubuntu started supporting
even third-party cloud computing platforms, such as Amazon EC2 [27].
36
Microsoft Office is a well-known application, which was released by Microsoft in 2008
[28]. It began with a basic word processor and some mathematical spreadsheet applications:
Microsoft Word, Microsoft Excel, and Microsoft PowerPoint. Later on, the ―
Pro‖ version, which
contains Microsoft Access and Schedule Plus, was released. This software works on both a
Windows OS and MAC OS. More than 35 languages are supported, and almost 80% of
enterprises have a version of it; 64% of enterprises maintain Office 2007. This software is
written in C++.
OpenOffice, commonly known as OOo, runs on different operating systems and is an
open-source software application [29]. It supports more than 110 languages and free software
that is written using a GUI toolkit. Its ISO/IEC standard [30] helps to open the document in
default format, and MS Office formats are supported as well. The latest version is 3.2, and
various components are available. Similar to Microsoft Office, a word processor called ―
Writer‖
is used for testing in this research.
Google Docs is a web-based spreadsheet, word processor, form, data storage, and
presentation service offered free of cost by Google [31]. The data storage capacity is around
1 GB, and it contains writer and spreadsheets within the same presentation program incorporated
by Tonic Systems. This allows users to create and edit documents in real time over the cloud. It
does not require any software or applications to be installed prior to use. Google Docs is the best
example of a ―
Software as a Service‖ model, and is offered on all recent browsers such as
Google Chrome, Safari, Firefox, and Internet Explorer. It also supports all proprietary formats
such as .doc, .xls, and even new versions like .docx, .xlsx, etc. A snapshot of Google Docs is
shown in Figure 22.
37
Figure 22: Snapshot of Google Docs
Microsoft Office Live is also a web-based service offered by Microsoft for sharing,
editing, and storing documents [32]. It supports 25 languages and is used for work, school, and
home projects for modifying data and sharing remotely without the help of a flash drive. A
document can be edited easily with the click of a button, where it opens up in Microsoft Office
and can be modified in real time. An application called Silverlight is a plug-in where users can
upload multiple documents. Different features of this service are online storage, software
compatibility, information sharing, resource and support, etc. Browsers like Internet Explorer,
Firefox, Safari, etc. support these services. Also, this can be used as an example for SaaS. It
permits user with a storage space of 5 GB of information.
4.2.2 Energy Measurement
In Ubuntu 9.04, the battery readings are obtained using the ―
state‖ of the battery present
in a file updated by an ACPI module [33]. In Windows 7, a lightweight battery meter, BatteryBar
[34], is used for obtaining the readings.
38
For Ubuntu, experiments were done using cloud and non-cloud applications like Google
Docs and Open Office, respectively. Various new interesting results have been obtained and are
explained in the next section. For Windows, this study had a wide scope due to the compatibility
of various applications. Cloud applications like Google Docs and Office Live and non-cloud
applications like Microsoft Office and Open Office were used.
The flow of packets for both operating systems was studied using ―
Wireshark.‖ They are
explained in the following sections under Windows and Ubuntu.
4.2.2.1 Windows
Google Docs was used for testing purposes. When ―t
yping‖ was performed, the packets
per second over time were captured and are shown in Figure 23. The number of packets per
second using Goggle Docs was greater than when using Ubuntu, which shows evidence of
considerable on-going communication.
Figure 23: Windows: Time vs. packets per sec
The packet delay for each of the captured packets is shown in Figure 24, which indicates
that the data transfer was continuous without delay between packets.
39
0.35
Time
0.3
0.25
0.2
0.15
0.1
0.05
1
1335
2669
4003
5337
6671
8005
9339
10673
12007
13341
14675
16009
17343
18677
20011
21345
22679
24013
25347
26681
28015
29349
30683
0
Figure 24: Windows: Packet sequence vs. delay (sec)
Packet lengths from the application were captured and mainly consisted of two packet
lengths, as shown in Figure 25, which indicates that the window length was fixed during the
communication.
30000
Number of packets
25000
20000
15000
10000
5000
0
Figure 25: Windows: Packet length vs. number of packets
40
4.2.2.2 Ubuntu
Google Docs was used for testing purposes to compare both operating systems. When
―
typing‖ was performed, the packets per second over time were captured and are shown in
Figure 26. The packets per second were less compared to that in Windows, and there were many
bursts in the communication.
Figure 26: Ubuntu: Time vs. packets per sec
The packet delay for each of the captured packets is shown in Figure 27, which indicates
that the data transfer is continuous without delay between packets
3.5
Time
3
2.5
2
1.5
1
0
1
464
927
1390
1853
2316
2779
3242
3705
4168
4631
5094
5557
6020
6483
6946
7409
7872
8335
8798
9261
9724
10…
10…
11…
11…
12…
12…
0.5
Figure 27: Ubuntu: Packet sequence vs. delay (sec)
41
The packet lengths from the application were captured and consist of varying packet
lengths, as shown in Figure 28, which indicates that the window length is not fixed during the
communication.
6000
Number of packets
5000
4000
3000
2000
1000
0
Figure 28: Ubuntu: Packet lengths vs. number of packets
On comparing both operating systems, Windows seemed to perform better
communication then Ubuntu. Later sections show their energy patterns for local and remote
computation.
4.2.3 Reproducibility
Independent of the external conditions, the energy values were able to be reproduced.
The experiments were repeated on the laptop with varying battery power over a similar time
interval throughout the research, but the average results obtained remain constant. In order to
obtain accurate results, experiments where repeated multiple times, but little change in the
energy patterns was found, which gives some confidence about the results.
42
4.3
Results
The results of this research study to run office applications remotely as well as locally are
subdivided in this section under Ubuntu 9 and Windows 7.
4.3.1 Ubuntu 9
As discussed in an earlier section, the cloud and non-cloud applications taken into
consideration were Google Docs and Open Office, respectively. ―
C Energy‖ and ―
NC Energy‖
were the energy levels in mWh consumed by Google Docs and Open Office applications,
respectively. For unbiased results, both were started at the same energy levels and for the same
time duration. For Google Docs, the Wi-Fi was ON, and minimum brightness and no graphics
were used. For Open Office, the Wi-Fi was OFF, and minimum brightness and no graphics were
used. The idle energy consumption over the similar conditions was also tabulated for a
comparison of both results. Graphs were plotted in MATLAB for the discussed scenarios and are
shown in Figure 29.
Figure 29: Ubuntu with cloud and non-cloud readings
43
4.3.2 Windows 7
Cloud-based applications and non-cloud-based applications for these scenarios were
C1 Energy‖
Google Docs and Office Live, and Microsoft Office and Open Office, respectively. ―
and ―
C2 Energy‖ were the energy levels in mWh consumed by Google Docs and Office Live,
NC2 Energy‖ were the energy levels consumed by Microsoft
respectively. ―
NC Energy‖ and ―
and Open Office, respectively. For Google Docs and Office Live, the Wi-Fi was ON, and
minimum brightness and no graphics were used. For Microsoft Office and Open Office, the WiFi was OFF, and minimum brightness and no graphics were used. The idle energy consumption
over the similar conditions was also tabulated for comparison with both results. Graphs were
plotted in MATLAB for the discussed scenarios and are shown in Figure 30.
Figure 30: Windows 7 with cloud and non-cloud readings
4.3.3 Comparison of Operating Systems
For comparison, the energy consumed in both versions is shown in Figure 31.
44
Figure 31: Cloud computing: Windows vs. Ubuntu
To obtain appropriate readings in both versions, the same data was used on all test
benches and was kept constant in the study. In the case of the idle state, the Ubuntu version
consumed 17.8% more than the Windows version. Cloud and non-cloud (local) versions in
Ubuntu took more than 23% and 30%, respectively.
4.4
Implications
From the graphs in Figure 31, it can be clearly seen that cloud computing is the energy-
saving scheme, as compared with non-cloud applications, while keeping the major energyconsuming components at their minimum levels during the experiments. In Ubuntu, even though
there is a close tradeoff between applications, the key factor that should be considered is that the
energy consumed by Wi-Fi also increases in the cloud scenario, which clearly violates the noncloud application part of this scenario.
4.5
Limitations
The accuracy of the experiments conducted was limited by various factors.
Measurements were taken according to the ―
state‖ of the battery, where there could be a
marginal error of +/- 3mA.
45
CHAPTER 5
CONCLUSION AND FUTURE WORK
The main objective of this research study was to run office applications remotely as well
as locally for different scenarios in a laptop. Results obtained in the case of office applications
prove that the cloud is more energy-efficient than local execution. This should help researchers
to understand the importance of developing more cloud applications in the years ahead, which in
turn may lower maintenance costs.
In the future, work over 3G, 4G, and netbook devices that study the implications on
energy consumption could be done. Also, cloud-based operating systems like Google’s
Chromium [35], Cloudo [36], iCUBE [37], eyeOS [38], and other new cloud OSs could be used
as testbeds.
46
REFERENCES
47
REFERENCES
[1]
―
What is Energy Efficiency,‖ Use Energy Efficiently: How to Use Energy Wisely at
Home and at school, URL: http://www.midamericanenergy.com/eew/efficiently/
efficiency.html [cited 21 August 2010].
[2]
―
Global Warming,‖ Climate Change 2001, URL:http://www.grida.no/publications/
other/ipcc_tar/?src=/climate/ipcc_tar/wg1/figspm-1.htm [cited 23 September 2010].
[3]
Pavel Somavat, Shraddha Jadhav, and Vinod Namboodiri, ―
Accounting for the Energy
Consumption of Personal Computing Including Portable Devices,‖ Portable Devices
Energy Consumption, URL: http://www.cs.wichita.edu/~vnambood/mypubs/eEnergy
09.pdf [cited 21 November 2009].
[4]
―
Cloud Computing,‖ Gartner Says Cloud Computing Will Be As Influential As EBusiness, URL: http://www.gartner.com/it/page.jsp?id=707508 [cited 24 September
2010].
[5]
―
IBM Lenovo SL400,‖ SL400, URL: http://www.retrevo.com/support/Lenovo-SL400Laptops-manual/id/21457ag116/t/2/ [citied 12 June 2009].
[6]
―
Ubuntu Operating System,‖ Ubuntu, URL: http://www.ubuntu.com/testing/jaunty/beta
[cited 10 July 2009].
[7]
―
Active Matrix Liquid Crystal Display,‖ LCD, URL: http://ieeexplore.ieee.org/stamp/
stamp.jsp?arnumber=01478049 [cited 11 June 2010].
[8]
―
Twisted Nematic Field Effect,‖ Nematic Crystal, URL: http://www.lcd-experts.net/
docs/ TN-Patent.pdf [cited 15 July 2010].
[9]
―
Warezone 2100—Gaming,‖ Warezone, URL:
2010].
[10]
―
Evony Online Strategy Gaming,‖ Evony, URL: http://www.evony.com/ [cited 9
February 2010].
[11]
―
Solitaire Microsoft Game,‖ Solitaire, URL: http://www.dmoz.org/Games/Card_Games
/Shedding_and_Accumulating/Solitaire/ [cited 10 August 2009].
[12]
―
IEEE standards,‖ IEEE, URL: http://www.ieee802.org/11/ [cited 1 September 2010].
[13]
―
Skype Software,‖ Skype, URL: http://about.skype.com/ [cited 3 August 2010].
[14]
―
SATA Cable,‖ Cable, URL: http://www.sata-io.org/ [cited 20 September 2010].
48
http://wz2100.net/ [cited 19 August
REFERENCES (continued)
[15]
―
Universal Serial Bus Port,‖ USB, URL: http://compnetworking.about.com/od/
usbnetworking/p/usb-port.htm [cited 14 April 2010].
[16]
―
BlueSoleil Bluetooth Dongle,‖ BlueSoleil, URL: http://www.bluesoleil.com/ [cited 11
June 2010].
[17]
Aqeel Mahesri and Vibhore Vardhan, ―
Power Consumption on a Modern Laptop,‖ Power
Consumption, URL: http://users.crhc.illinois.edu/mahesri/classes/project_report_cs497
yyz.pdf [cited 21 August 2009].
[18]
Judith Hurwitz, Robin Bloor, Marcia Kaufman, and Fern Halper, ―
Cloud Computing for
Dummies,‖ in Cloud Computing, Wiley Publication, Inc., 2010.
[19]
Ardalan Amiri Sani, Hasan Dumanli, Lin Zhong, and Ashutosh Sabharwal, ―
PowerEfficient Directional Wireless Communication on Small Form-Factor Mobile Devices,‖
Power Efficient Directional Wireless Communication, URL: http://portal.acm.org
/citation.cfm?id=1840845.1840904 [cited 11 January 2010].
[20]
Jiayang Liu and Lin Zhong, ―
Micro Power Management of Active 802.11 Interfaces.‖
Micro Power Management, URL: http://www.ruf.rice.edu/~mobile/publications
/liu08mobisys.pdf [cited 12 February 2010].
[21]
Hady S. Abdelsalam, Kurt Maly,Ravi Mukkamala, Mohammad Zubair, and David
Kaminsky, ‖Analysis of Energy Efficiency in Clouds,‖ Energy Efficiency, URL:
http://www.computer.org/portal/web/csdl/doi/10.1109/ComputationWorld.2009.38 [cited
24 July 2009].
[22]
View of power efficiency on mobile
Marius Marcu, Dacian Tudor, and Sebastian Fuicu, ―
Applications,‖ Power Efficiency, URL: http://www.springerlink.com/content/h64k3j505
k7156m1/ [cited 10 April 2008].
[23]
Energy Efficiency Cloud
Andreas Berl, Erol Gelenbe, and Marcodi Girolamo, ―
Computing,‖ Energy Efficiency, URL: http://comjnl.oxfordjournals.org/content/
53/7/1045.abstract [cited 19 August 2009].
[24]
P. Simoens, B. Vankeirsbilck, and F. AzmatAli, ―
Characterization of Power
Consumption in Thin Clients Due to Protocol Data Transmission over IEEE 802.11,‖
Power Consumption in Thin Clients Due to Protocol Data Transmission, URL:
http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=5291601&tag=1[cited 21 March 2010].
[25]
―
Office Live and Google Docs,‖ Features Comparison, URL: http://www.rea
dwriteweb.com/archives/office_live_workspace_vs_google_docs_feature_by_feature.php
[cited 10 August 2009].
49
REFERENCES (continued)
[26]
―
Windows 7 Operating System,‖ Windows 7, URL: http://windows.microsoft.com/enus/windows7/products/home?os=nonwin7 [cited 12 August 2007].
[27]
―
Amazon Elastic Compute Cloud,‖ Amazon EC2, URL: http://aws.amazon.com/ec2/
[cited 16 August 2007].
[28]
―
Microsoft Office Suite,‖ MS Office, URL: http://office.microsoft.com/en-us/ [cited 20
August 2010].
[29]
―
Open Office Suite,‖ Open Office, URL: http://about.openoffice.org/ [cited 12 September
2010].
[30]
―
ISO/IEC Standard,‖ ISO, URL: http://www.standardsinfo.net/info/livelink/fetch/
2000/148478/6301438/index.html [cited 10 April 2008].
[31]
―
Google Docs Online Software,‖ Google Docs, URL: http://www.google.com/google-ds/tour1.html [cited 21 September 2010].
[32]
―
Microsoft Office Live Beta,‖ Office Live, URL: http://www.officelive.com/en-us/ [cited
15 September 2010].
[33]
―
About ACPI Battery,‖ ACPI, URL: https://help.ubuntu.com/community/ACPIBattery
[cited 10 November 2009].
[34]
―
Osiris Development—BatteryBar,‖ Battery Measurement, URL: http://osirisdevelo
pment.com/BatteryBar/ [cited 9 December 2008].
[35]
―
Google Chromium Operating System,‖ Chromium, URL: http://www.chromium.org/
chromium-os [cited 14 November 2009].
[36]
―
Cloudo Operating System,‖ Cloudo, URL: http://beta.cloudo.com/ [cited 10 August
2010].
[37]
―
iCUBE Operating System,‖ iCUBE, URL: http://www.oos.cc/login.html [cited 12
January 2010].
[38]
―
eyeOS,‖ eyeOS, URL: http://beta.my.eyeos.org/ [cited 15 August 2010].
50
APPENDIX
51
APPENDIX
MATLAB CODE
For Ubuntu
excel_file = 'ubuntu.xls';
var1 = xlsread(excel_file);
a= var1(:,1);
b1= var1(:,2);
b2= var1(:,3);
b3= var1(:,4);
lplot(a,[b1,b2,b3])
c= legend('Idle','Cloud(Google Docs)','Noncloud(Open Office)',3);
xlabel('Time in mins');
ylabel('Energy in mWh');
grid on;
For Windows
excel_name = 'windows.xls';
excel_file = xlsread(excel_name);
a= var1(:,1);
b1= var1(:,2);
b2= var1(:,3);
b3= var1(:,4);
b4= var1(:,5);
lplot(a,[b1,b2,b3,b4])
c= legend('Idle','Cloud(Google Docs)','Cloud(Office Live)','Noncloud(Microsoft Office)',4);
52
APPENDIX (continued)
xlabel('Time in mins');
ylabel('Energy in mWh');
grid on;
53