A Method for Characterizing Energy Consumption in Android Smartphones

idleheadedceleryMobile - Wireless

Dec 10, 2013 (3 years and 7 months ago)

50 views

A Method for Characterizing Energy Consumption
in Android
Smartphones

Authors:

Luis Corral,
Anton B. Georgiev
, Alberto Sillitti
, Giancarlo Succi

Center
for Applied Software
Engineering, Free
University of Bozen/
Bolzano,
Bolzano
,
Italy




GREENS 2013, ICSE 2013

San Francisco, CA, USA

20.05,2013

www.en
-
act.eu

Motivation


Smartphones have a market penetration in
millions

both in
active users
and
sold devices
.


Power autonomy
requirements

of mobile
devices with rapidly
increasing computing capabilities.


Reduce battery garbage.


In order to
save

energy first
we need to
measure

it.


L. Corral,
A. Georgiev
, A. Sillitti, G. Succi | Center for Applied Software Engineering |
Free University of Bozen/Bolzano

Our approach


We aim to characterize the energy consumption based on the
time that battery
lasts.


relate
the status of the components of the smartphone with the
battery
level


collect
available sensors'
data surveying parameters exposed by OS


CharM

-

Android application tool for collecting energy related OS
parameters of the device component.


Experiment:


Setup
-

Three devices with different hardware and software
configurations (
HTC Nexus One
,
Samsung Galaxy
,
Nexus
7
tablet
,)


Data
Collection


Three stages with multiple
2 hours

long iterations:


Normal Mode


daily usage of the device, serve as a baseline.


Airplane Mode


all the
communication interfaces
disabled.


Component stress Mode


every iteration one of the components
was put in high power consumption state, while the others were
disabled.




L. Corral,
A. Georgiev
, A. Sillitti, G. Succi | Center for Applied Software Engineering |
Free University of Bozen/Bolzano

Results

L. Corral,
A. Georgiev
, A. Sillitti, G. Succi | Center for Applied Software Engineering |
Free University of Bozen/Bolzano

Execution

Mode

Nexus
One

Nexus 7

Galaxy
Nexus

Normal mode

10%

2%

4%

Airplane mode

2%

0.2%

1%

CPU stress

40%

17%

26%

OLED stress

35%

24%

20%

Video playback
stress

12%

7%

10%

WiFi

stress

24%

22%

32%

GPS stress

17%

10%

15%

Percentage of Battery Discharge in 2
Hours

Component

Impact to discharge
cycle (average)

1. CPU stress

27.66%

2. OLED stress

26.33%

3.
WiFi

interface

26.00%

4. GPS interface

14.00%

5. Video

playback

9.60%

6. Normal mode

5.33%

7. Airplane mode

1.06%







500+ hours of data collection.


The
most energy hungry
components are
CPU
,
OLED

and
WIFI

interface.

Ranking
of Energy Hungry Components

Challenges


Improve
CharM:


Add
more parameters
that can be taken from the OS (e.g. battery
temperature
).


Provide enhanced
reporting features
(e.g. customized plots, tables
and reports
).


Improve collected
data

accuracy

and
completeness
.


Green metrics?


Green security?


Testing for energy efficiency?


L. Corral,
A. Georgiev
, A. Sillitti, G. Succi | Center for Applied Software Engineering |
Free University of Bozen/Bolzano

Questions?

L. Corral, A. Georgiev, A. Sillitti, G. Succi | Center for Applied Software Engineering |
Free University of Bozen/Bolzano