Decision to offload the task to Cloud for increasing energy efficiency of Mobile Phones

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395 -0056
Volume: 04 Issue: 01 | Jan -2017 www.irjet.net p-ISSN: 2395-0072
© 2017, IRJET | Impact Factor value: 5.181 | ISO 9001:2008 Certified Journal | Page 762
Decision to offload the task to Cloud for increasing energy efficiency
of Mobile Phones
Hiren Dand1, Gulabchand K. Gupta2
1Research Scholar, Department of Computer Science and Engineering, Shri JJT University, Jhunjhunu, Rajasthan,
India. dandhiren@yahoo.co.in
2 Department of Computer Science and Engineering, Shri JJT University, Jhunjhunu, Rajasthan, India.
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract -Smartphones havebecomeanintegralpartofour
life. They provide a variety of services including those required
for daily use. Battery life of smart phones is limited and the
main hindrance in its utility. Energy efficiencyofbatterycanbe
increased by offloading someof the tasks to cloud.Thedecision
to offload the task is crucial and dependentonmanyfactors.To
benefit from task offloading,theenergyconsumedinoffloading
activities need to be estimatedand the decision canbetakenas
to whether to offload the task to the cloud or to perform it
locally. The paper presents the comprehensive way of energy
estimation and decide whether to offload the task to cloud for
increasing energy efficiency of mobile phones.
Key Words: Mobile Cloud Computing, Smartphones,
Energy estimation, offloading decision.
1. INTRODUCTION
The battery power, processing power and memory of the
smartphones is limited. In the last few years there has been
tremendous advancement in the phone batteries. From
Nickel Cadmium batteries, which suffered from memory
effect to Nickel Metal Hydride with high cost to today’s
Lithium Poly ion batteries with no memory effect and light
weight, the batteries have come a long way. [1] Smartphones
today come with powerful operating systems like Windows,
Android, Blackberry, Apple iOS and Symbian. They are
capable of running applications similar to the onethatrunon
desktop computers. These applications and other
smartphone features consume energy that hinders theuseof
smartphones.
There is a need to reduce energy consumption and a
number of researchers are working towards it. Many
techniques are suggested like smart batteries, power
scheduling, increasing efficiency of operating systems and
applications, energy-aware communication protocols and
task offloading. Task offloading is a favorable technique for
reducing energy consumption with the development of the
high speed wireless Internet access. High speed networks
increase the connection availability between the mobile and
the cloud.[10][11] Using the offloading technique,
smartphones shave their energy by offloading heavy
computation tasks to the cloud. [6][9][21] The mobile device
offloads the heavy task to the cloud, the cloud executes the
tasks and send the results back to the mobiledevice.Thiswill
enable the mobile device to save the energy spent in
executing the task. An example of a task could be video
format conversion, in which the mobile uploads the video to
the cloud, the cloud converts it into desired format fitting the
smartphone capabilities. The processing will take place on
the cloud.
Task offloading is a crucial technique as in some cases it may
increase the energy consumptionofsmartphones.Everytask
involves data and processing.
In this work, we prove that the energy efficiency of the
mobile phone can be improved by offloading the tasks to
cloud.
Though there can be different types of connectivity, WLAN is
used for modelling.
2. RELATED WORK
Several techniques for offloading have been proposed. The
techniques can be categorized into three methods based on
the type of the remote machine. [8] [17] [20]
The first technique uses a web proxy. The web proxy lies
between the web server and the mobile device. The mobile
device sends the request to the web proxy and thewebproxy
forwards the request to the web server. The web server
processes the request and delivers the processed content to
the proxy which in turn delivers the content to the mobile
device. [7]
The second technique involves offloading the task to a local
high performance server. The server and the mobile device
are located in the sameor nearbynetwork.Themobiledevice
would offload the heavy computation task to the server, the
server will process and generate the results. The mobile
device would download the results. [9] [12]
The third technique involves offloading the tasks to cloud.
The cloud provides different resources to the mobile device
like storage and processing. [13]
In this paper, weuse the third technique and decide whether
it is feasible to offload a particular task or local processing is
preferable.

3. THE PROPOSED MODEL
The proposed model consists of the mobile device and the
cloud both of which are connectedtotheInternetasshownin
figure 1. The mobile devices are connected to the Internet
through the WLAN access point. They could be connected
through Base StationSubsystem(3G/4G).Themobiledevices
provide mobile computing facilities to the end users via
different apps. The cloudconsistsoftheclouddatacenterand
the cloud service provider, which can be accessed through
the Internet. The cloud provides the mobile devices with all
the functionalities needed for mobile computing and the
processing facilities for the offloaded tasks. [2]
The mobile devices access the cloud via the Internet.
Offloading is hence dependent on the network. The network
interface cards have their own characteristics. One of the
important characteristic is the data rate supported by NICs.
Http protocol will be used for offloading. The wireless NICs
and the protocols are the most important factors that affect
the cost of offloading the task.[14][16]
Virtualization is a fundamental featureof cloudcomputing.It
allows the applications from different users to run on
different virtual machines. This provides separation and
protection. [15]
MobilePhones[3]
WLAN
Internet
Cloud
Fig-1: The Proposed Model
4. ENERGY ANALYSIS FOR COMPUTATON
OFFLOADING
Suppose a particular task can be processed in I instructions.
Let Sc be the speed of the cloud server and Sm be the speed of
the mobile device.
Time to complete task on cloud =
Time to complete the task on mobile device =
Let L be the number of bytes exchanged between the mobile
device and the cloud. Let the network bandwidth be B.
Time taken to transmit and receive data =
Let Pm be the power consumed by mobile device for
processing. Let Pi be the power consumed by mobile device
while it is idle. Let Pt be the power consumed by mobile for
transmitting data and let Pr be the power consumed by
mobile for received data.
If the processing is done on the mobile device, the energy
consumption will be . If the processing is done
on the cloud, the energy consumption would be
The amount of energy saved would be
.…. (1)
Suppose that the processing on cloud in N times faster
than on the mobile device, then Equation(1)can
be rewritten as
….. (2)
When the result of equation (2) is positive, energy is saved.
The result will be positive if is small compared to and N
is sufficiently large.
5. ANALYTICAL MODEL USING WLAN
Most of the smartphones and mobile devices support
802.11g network. We consider 802.11gsinglechannel Wi-Fi
network. It uses CSMA/CA protocol. If the mobile device
needs to transmit a data packet, it senses the channel. If the
channel is idle for DIFS duration, the device transmits RTS
packet. If the channel is busy, the mobile device defers the
transmission. It detects idle DIFS and waits for randomback
off time to avoid collision. The random back off delay is
chosen in the range where W is called back off
window or contention window (CW). The initial CW is set to
Internet Service Provider
Cloud Data Center and Cloud
Service Provider

W = 32. The value of the back off timer is decreased as long
as the medium is sensed to be idle for a DIFS and stopped
when a transmission is detected on the medium and
resumed when the channel is detected as idle again for a
DIFS interval. When the back off reaches0,themobiledevice
transmits if packet. In IEEE 802.11, time is slotted in a basic
time unit, which is the time needed to detect the
transmission of a packet from any other station. If two or
more mobile devices decrease their back off timer to 0atthe
same time, collision occurs and CW is doubled for each
retransmission until it reaches maximum value. [4][18][19]
SIFS is used to give priority access to the ACK packets. When
the packet is received correctly, the receiver waits for SIFS
interval immediately after the reception is completed and
transmits an ACK back to the source mobile device to
confirm the reception. If the source device dos not receive
and ACK due to collision or transmission errors, it
reactivates the back off timer after the channel remains idle
for EIFS interval.
Assume one mobile device is communicating with an Access
Point using TCP (e.g. transferring a file via FTP, accessing a
web page via HTTP). Further assume that each TCP data
packet is followed by TCP ACK packet. To transfer the data
segment there will be [2][4][6][8]:
a. Silence during at least one DIFS slot, signaling that
medium is available. (This could be more than one if
back off is being executed.)
b. The data frame containing TCP data.
c. The SIFS gap between data frame and 802.11 ACK
frame.
d. The 802.11 ACK frame.
To transfer TCP ACK segment there will be:
a. Silence during at least one DIFS slot, signaling that
medium is available. (This could be more than one if
back off is being executed.)
b. The data frame containing TCP ACK.
c. The SIFS gap between data frame and 802.11 ACK
frame.
d. The 802.11 ACK frame.
In addition to the payload data, thedata framehasadditional
36 bytes of data (28 bytes of 802.11 MAC header for various
control and management, error detection and addressing, 8
bytes header to identify the network layer protocol.)
To transfer a payload of 1460 bytes the packet size is 1460
bytes (payload) + 20 bytes (TCP header) + 20 bytes (IP
header) = 1500 bytes + 28 bytes (802.11 MAC header) + 8
bytes (network layer identification) = 1536 bytes.
For TCP ACK segment of 40 bytes the total packet size is 40
bytes + 28 bytes (802.11 MAC header) + 8 bytes (network
layer identification) = 76 bytes.
5.1. Maximum throughput of 802.11g:
SIFS = 10 µs
Short Slot time (α) = 9 µs
Long Slot time (β) = 20 µs
DIFS = 2 * Slot time + SIFS = 28 µs
Preamble = 20 µs
Signal extension = 6 µs
Data rate = 0.25 MSymbols/s
Size of each symbol = 216 bits
Data rate in Mbps = 216 * 0.25 = 54 Mbps
Time to transmit each symbol= 4 µs
Size of TCP data packet = 1536 bytes
= 12288 bits
Symbols needed to transmit
TCP Data Packet = 12288/216  57
Size of 802.11 ACK packet = 14 bytes
= 112 bits
802.11 ACK packet = 112/216  1
Size of TCP ACK packet = 76 bytes
= 608 bytes
TCP ACK packet = 608/216  3
If the network is only 802.11g, we can use Short time
slots.
Time required to transmit TCP data packet
= DIFS + 802.11 data + SIFS + 802.11 ACK
= 28 µs + (20 µs + 57 × 4 µs + 6 µs) + 10 µs + (20 µs
+ 1 × 4 µs + 6 µs)
= 28 µs + 254 µs + 10 µs + 30 µs
= 322 µs
Time required to transmit TCP ACK packet
= DIFS + TCP ACK data + SIFS + 802.11 ACK
= 28 µs + (20 µs + 3 × 4 µs + 6 µs) + 10 µs
+ (20 µs + 1 × 4 µs + 6 µs)
= 28 µs + 38 µs + 10 µs + 30 µs
= 106 µs
Total time to transmit 1460 bytes = 322 µs + 106 µs
= 428 µs ……. (3)
Throughput
Similarly, the throughput of 802.11b, 802.11g(CTS to SELF)
and 802.11g (RTS – CTS) is calculated and the values are as
follows:
Maximum throughput of 802.11b is 5.6 Mbps
Maximum throughput of 802.11g (CTS to Self) is 13 Mbps
Maximum throughput of 802.11g (RTS - CTS) is 9 Mbps

6. EXPERIMENTAL SETUP
The experimental setup consists ofsmartphone(Usedhereis
Redmi 1S) with power tutor software, video converter
software, the cloudservice(simulatedonthecomputers),Wi-
Fi router connected to the Internet and laptop.
The file used is 30 MB flv 720p converted to mp4 format of
size 11 MB with the quality for mobile 320 X 240 pixels.
Battery monitor is used to calculate the energy used. The
tasks involved are:
T1: The file is kept on the mobile phone and converted from
flv to mp4 using Video Converter for Android.
T2: The file is kept on the mobile phone and converted from
flv to mp4 by sending the file to the cloud, converting the file
on the cloud and downloading the converted file on the
mobile again.
Fig. 2: Experimental Setup
T3: The flv file is kept on the cloud and request is sent from
the mobile to download the file. The conversion of the file
takes place on the mobile.
T4: The flv file is kept on the cloud and request is sent from
the mobile to convert the file in the mp4 format and
download the file in the mp4 format. The conversion of the
file takes place on the cloud.
The following table summarises the time taken and the
energy used for the four tasks:
Table -1: TASKS, TIME TAKEN and ENERGY USED
Task
Type
Residing / Carried out at
Data Processing
Time
Taken
Energy
Used
T1 Local Local 180 s 31 J
T2 Local Cloud 120 s 21 J
T3 Cloud Local 238 s 44 J
T4 Cloud Cloud 42 s 9.6 J
7. RESULTS AND DISCUSSION
Table 1 summarises the results.
T1: When the data is on themobile phone and theprocessing
is also on the mobile phone, the time taken is 180 s and the
energy consumed is 31 J.
T2: When the data is on themobile phone and theprocessing
is to be done on the cloud, the task is broken into three steps:
Step1: Upload the data on the cloud (30 MB flv file was
uploaded, the timetakenwas63secondsandtheenergyused
was 15 J.
Step 2: The file was converted from flv to mp4 format on the
cloud. The time taken was 36 seconds and the mobile phone
was idle during this time consuming 1 J of energy.
Step 3: Downloading the file in mp4 format. The size of the
file to be downloaded was 11 MB. The time taken do
download was 21 seconds and the energy consumed was 5 J
Total Time Taken = 63 + 36 + 21 = 120 s
Total Energy Used: 15 + 1 + 5 = 21 J
T3: The data resides on the cloud. It is downloaded (30 MB
flv file). The processing is done on the mobile itself. The task
differs from T1 in that, it involves additional step of
downloading the flv file on 30 MB.
The time taken to download the file is 58 seconds and the
energy consumed is 13 J.
The conversion thentakesplaceonthemobilein180seconds
consuming 31 J of energy.
Total Time Taken = 58 + 180 = 238 seconds
Total Energy Used = 13 + 31 = 44 J
T4: Here, we only need to play the mp4 file directly from the
cloud. A request is sent to the cloud to convert the file. The
request takes just few milliseconds (300). The file is
converted in 36 s and during this time the mobile is idle
consuming 1 J of energy. The file is then played directly from
the cloud. The duration of the video is 2.38 minutes (158
seconds). The energy consumed in viewing the file online is
just 8 J. he time taken to send the request to play the file is
gain only 300 ms and the requestwasservicedin5.4seconds.
The energy used is sending the requests is 0.3 + 0.6 = 0.6 J
Total Time Taken = 0.300 + 36 +0.300 + 5.4 = 42 seconds
Total Energy Used = 0.6 + 1 + 8 = 9.6 J
The results obtained are in accordance with the analytical
model.
The decision to offload the task to the cloud depends on
several factors:
 The size of the file to be processed. If the size of the
file is large, the processing needs to be done on the
cloud.

 The amount of computation needed to obtain the
results. If the computation is heavy, the processing
needs to be done in cloud.
 The amount of battery available. If the battery
available is more than 25%, then the task can be
offloaded to the cloud.
 The available bandwidth. The bandwidth plays a
vital role in offloading. Greater the bandwidth, the
faster to offload the task.
If the conditions are not met, local processing would be
the only solution.
8. CONCLUSION AND FUTURE SCOPE
The energy efficiency of the mobile phone can be increased
by offloading the task to cloud. We observe from the results;
the speed of execution is faster and energy consumption is
lower when the task is offloaded to the cloud. It is vital to
compute the energy that will be consumed in processing.
The offloading is advantageous only if the energy consumed
in offloading the task is less than the energy consumed
without it. The IEEE 802.11g standard was usedforanalysis.
The analysis and the experiments can be carried out for
newer IEEE802.11x standards and 3G and 4G interfaces. In
future, the analysis can be extended to the newer 802.11xx
standards.
The future mobile phones can be designed with built in
software to take the decision. Whenever any task is to be
executed, the software will check the conditions and
constraints and decide to offload or execute locally. The
mobile service operators can have their clouds for the
subscribers and the service can be provided at the minimal
cost.
REFERENCES
[1] A. P. Miettinen and J. K. Nurminen, “Energy efficiency of
mobile clients in cloud computing,” in Proc. 2010
USENIX Conference on Hot Topics in Cloud Computing.
[2] WLAN: Wikipedia.
https://en.wikipedia.org/wiki/Wireless_LAN.
[3] https://www.google.co.in/url?sa=i&rct=j&q=&esrc=s&s
ource=images&cd=&cad=rja&uact=8&ved=0ahUKEwjfp
bzezprRAhXFrY8KHdezAxkQjRwIBw&url=http%3A%2F
%2Fblog.quikr.com%2F2014%2F04%2F27%2Fhow-to-
keep-mobile-phone-in-good-
condition%2F&psig=AFQjCNFu2ZB7Imwp1CuGSuPzlW
DtYtzhuw&ust=1483142421703740
[4] E. D. Demaine, M. Ghodsi, M. T. Hajiaghayi, A. S. Sayedi-
Roshkhar, and M. Zadimoghaddam. Scheduling to
minimize gaps and power consumption. In SPAA ’07:
Proceedings of the nineteenth annual ACM symposiumon
Parallel algorithms and architectures, pages46–54,New
York, NY, USA, 2007. ACM.
[5] Z,Jiang and L.. Kleinrock Web prefetching in a mobile
environment.In IEEEPersonal Communications,volume
5, pages 25–34, September, 1998.
[6] Y. Xiao, R. S. Kalyanaraman, and A. Yla-Jaaski. Energy
consumption of mobile youtube: Quantitative
measurement and analysis. In NGMAST ’08: Proceedings
of the 2008 The Second International Conference on Next
Generation Mobile Applications, Services, and
Technologies, pages 61–69, Washington, DC, USA, 2008.
IEEE Computer Society.
[7] Yang, K., Ou, S., Chen H-W. (2008). On Effective
Offloading Services for Resource Constrained Mobile
Devices Running Heavier Mobile Internet Applications.
IEEE Communications Magazine, January 2008
[8] K. Naik, ‘‘A survey of software based energy saving
methodologies for handheld wireless communication
devices,’’ Dept. Electron. Commun. Eng., Univ. Waterloo,
Waterloo, ON, Canada, Tech. Rep. 2010-13, 2010.
[9] X. Gu, K. Nahrstedt, A. Messer, I. Greenberg, and D.
Milojicic. Adaptive Ofﬂoading Inference for Delivering
Applications in Pervasive Computing Environments. In
Proceedings of the First IEEE International Conference
on Pervasive Computing and Communications
(PerCom), 2003.
[10] http://www2.alcatel-
lucent.com/blogs/techzine/2010/mobile-cloud-
computing-challenges/
[11] Kumar K, Lu YH (2010) Cloud computing for mobile
users:can offloading computation save energy? IEEE
Comput 43(4):51–56
[12] N. Fernando, S. W. Loke, and W. Rahayu, “Mobile cloud
computing: a survey”, Future Gener. Comp. Syst.,vol.29,
no. 1, pp. 84–106, 2013.
[13] Khan, M. Othman, S. Madani, and S. Khan, "A survey of
mobile cloud computing application models," IEEE
Communications Surveys & Tutorials, issue 99, 2013.
[14] B.-G. Chun, S. Ihm, P. Maniatis, and M. Naik, “Clonecloud:
boosting mobile device applicationsthroughcloudclone
execution,” arXiv preprint arXiv:1009.3088, 2010.
[15] H. T. Dinh et al., “A Survey of Mobile Cloud Computing:
Architecture, Applications, and Approaches,” Wireless
Commun. and Mobile Computing, Oct 2011.
[16] E. Marinelli. Hyrax: Cloud computing on mobile devices
using mapreduce. Master’s thesis, Computer Science
Dept., CMU, September 2009.
[17] SILVEN, O., AND JYRKK ¨ A, K. Observations on power-
efﬁciency trends in mobile communication devices.
EURASIP J. Embedded Syst. 2007, 1 (2007), 17–17.
[18] Krashinsky R, BalakrishnanH(2002)Minimizingenergy
for wireless web access with bounded slowdown. In:
Proceedings of the 8th annual international conference
on Mobile computing and networking.ACM,pp119–130
[19] G. Anastasi, M. Conti, M. Di Francesco, and A. Passarella.
Energy conservation in wireless sensor networks: A
survey. Ad Hoc Networks, 7(3):537–568, 2009.
[20] Z. Zhuang, K.-H. Kim, and J. P. Singh. Improving energy
efciency of location sensing on smartphones. In
Conference on Mobile Systems, Applications, and
Services (MobiSys), 2010.
[21] Cuervo, E., Balasubramanian, A., Cho, D.-k., Wolman, A.,
Saroiu, S., Chandra, R., Bahl, P.: MAUI: making
smartphones last longer with code offload. In
InternationalConference on Mobile Systems,
Applications and Services,pp. 49-62

Decision to offload the task to Cloud for increasing energy efficiency of Mobile Phones

More Related Content

What's hot

What's hot (18)

Similar to Decision to offload the task to Cloud for increasing energy efficiency of Mobile Phones

Similar to Decision to offload the task to Cloud for increasing energy efficiency of Mobile Phones (20)

More from IRJET Journal

More from IRJET Journal (20)

Recently uploaded

Recently uploaded (20)

Decision to offload the task to Cloud for increasing energy efficiency of Mobile Phones