Performance Comparison of Android Messengers

Rama Bhatia & Renu Bagoria
International Journal of Computer Science & Security (IJCSS), Volume (13) : Issue (2) : 2019 24
Performance Comparison of Android Messengers
Rama Bhatia rmabhatia@gmail.com
Department of Computer Science and Engineering
JaganNath University
Jaipur, India
Lecturer, School of Engineering and Technology,
PSB Academy, Singapore,
Renu Bagoria renu.bagoria@jagannathuniversity.org
Department of Computer Science and Engineering
JaganNath University
Jaipur, India
Abstract
The increasing demand of Android applications compels the developers to think and develop
applications with efficient use of memory usage, CPU utilisation and UI rendering speed. The
literature survey reveals that very few works have been reported for measuring the performance
of various android applications. In this paper the performance analysis of most of the popular
mobile applications have been carried over using Eclipse with SDK Tools and Android Virtual
device. CPU utilisation, Memory usage and User Interface (UI) rendering speed have been
considered as the performance metric. The performance of most commonly used apps like
WhatsApp, Viber, WeChat & Imo have been analysed. The experimental result shows that
(i) Utilisation of CPU in case of Wechat is more as compared with WhatsApp, Viber, & Imo.
(ii) In case of Memory Usage, java.lang.string class found as a major problem suspect for
memory leak problem.
(iii) And for Rendering Speed in case of viber is taking lesser time as compared with
WhatsApp,
Viber, & Imo .
Keywords: Android App, WhatsApp, WeChat, Viber , Imo, Performance Analysis.
1. INTRODUCTION
Due to the increasing demand of Android devices and the various mobile applications, the
performance assessment of various android apps is required. Mostly, Android apps are easily
downloadable from GooglePlay [17] and the no of apps are increasing significantly. WhatsApp,
WeChat, Imo and Viber are the most popular messaging apps in present days. According to data
from statista (April 2016), there are around 1,000 million monthly active users in WhatsApp and is
the most popular messaging app while Viber, Line, WeChat and Imo are the other popular
messaging apps. The key to success of any messaging apps is depending on many performance
factors[1]. From developers point of view, a good messaging-app is that which effectively utilises
CPU, memory and consumes less energy. This motivate us to further asses the performance
efficiency of the mobile chat messenger applications in terms of resource utilization.
2. RELATED WORK
Liu Pu[19](2009) discussed the Short Message Service Architecture and their services. Anthony
Gutierrez et al [20] (2011) developed BBench to assess web-browser’s rendering performance.
Dhinakaran Pandiyan et al.[21](2013) developed a mobile platform benchmark suite,
MobileBench and presented the performance and energy characterizations for Mobile Bench.

Thiago Soares Fernandes et al.[18](2014) evaluated the CPU and memory of some android
application. Linares- Vasquez et al. [2](2015) studied the various practices and tools used by
Android developers to detect and fix the performance. Moreover, it was found that for
performance assessment of CPU Traceview[7] is the top most used tool followed by MAT and
then DDMS, Systrace and Logcat are taking the next place [2]. Many others are also suggesting
to use different SDK tools such as DDMS[4], MAT[3][5], Systrace[6][7], Hierarchy Viewer[8] for
performance analysis on different parameters of android apps [9]. In this paper, the performance
of WhatsApp, WeChat, Viber and Imo(the top most used android mobile chat messengers[9])has
been analysed using the Profiling Software - Eclipse with SDK Tools ( the known tools for
performance analysis [2][3][5][8][6][7][8])and Android Virtual device(Nexsus Emulator). The
experimental results provide the resource utilisation of various modules of the messaging apps.
Comparative Research work is published [22].
3. ANALYSIS CRITERIA AND MEASURING TOOLS
This section first gives you an overview of the parameters that are computed and the rest section
continues with the tools and methods used to perform the calculation are given.
3.1 Analysis Criteria
This section focussed the methods and techniques used for evaluating the performance. All the
experiments are done on the standard APIs of the mentioned messengers on android platforms
[13].
3.1.1 CPU Utilization- If an application takes more CPU time then it will impact adversely on
other processes running on the device. In the proposed work, the following four common modules
have been taken for measuring the CPU utilisation time as these are the commonly used
modules in all the apps.
(i) android.os.handler.dispatchMessage
(ii) android.os.handler.handleCallback
(iii) android.view.View.layout
(iv) android.view.ViewGroup.layout
Initially the evaluation has been done by restricting the no of contacts to three and finally the
evaluation is done by increasing the contacts to thirty.
During the initial stage of the application i.e. when the application initially started, we send the
messages and use the call procedure within 90,000 msec. This is the common evaluation criteria
used in all the applications.
3.1.2 Memory Utilization- Memory utilization is the total usage of memory used by the processes
which Includes allocated Heap, Free memory and percentage of used memory. Memory leakage
is the one of the major constraint which degrades the performance of any application. Hence
monitoring of memory utilization is one of the important step for the developers to check the
instability of the application due to the memory leakage. In addition to this, garbage collection is
also an important criteria for efficient use of memory. This parameter is evaluated using repeated
Garbage Collection (GC) calls during application launching state.
3.1.3 UI Rendering Speed- Good UI of any application is one of the key parameter for success
of any application. The bad design of the application layout interrupts the loading speed of the
application. To evaluate this parameter, the following four criteria were considered:
(i) No. of views of the application
(ii) Rendering speed parameters like Measure time(ms), Layout time (ms) and Draw
time(ms) [8].
3.2 Measuring Tools
In the proposed work, Eclipse is used extensively for evaluation of all the parameters. CPU
utilisation has been measured by using Traceview whereas for Memory measurement, Eclipse

Memory Analyzer (MAT), Heap and HPROF has been used. Hierarchy Viewer has been used for
measuring UI rendering speed. The advantage of using Traceview for measuring the CPU is that
it gives an graphical presentation of CPU utilisation time [15]. Similarly the advantage of Eclipse
Memory Analyzer (MAT) is that it gives a detailed analysis of used memory and memory
leaks[16]. In addition to this it also generates the memory leak suspects report. For
measurements of UI rendering speed, Hierarchy Viewer[11]is selected as it visualize the nested
behaviour of screen views and the tool also finds the flaws in design layout. In addition to the
above mentioned tools, Nexsus_S, Android 6 has been used as an emulator.
4. RESULTS AND DISCUSSION
In this section, the quantitative assessment of the mentioned messenging applications with
respect to CPU Utilization, Memory usage and Rendering Speed have been presented.
4.1 CPU Utilization
In this paper, the CPU Utilization of different modules is measured on the basis of load factors (α)
where α represents number of contacts in an application.
The following modules have been considered:
4.1.1 android.os.handler.dispatchMessage
TABLE 1: CPU Utilization of android.os.handler.dispatchMessage.
From the above table 1 it is observed CPU Utilization in case of Wechat is more as compared to
WhatsApp, Viber and Imo.
4.1.2 android.os.handler.handleCallback
TABLE 2: CPU Utilization of android.os.handler.handleCallback.
From the above table 2, it is observed that in case of Callback module, CPU Utilization in case of
WeChat is more as compared to WhatsApp, Viber and Imo.
α =30α =3
Inclusive
CPU Time%
Inclusive CPU Time
(ms)
Inclusive CPU
Time%
Inclusive CPU Time
(ms)Application
59.290065.1410WhatsApp
45.887061.5320Viber
70.21002070.310290WeChat
70.5600068.16520Imo
α =30α =3
Inclusive CPU
Time%
Inclusive CPU Time
(ms)
Inclusive CPU
Time%
Inclusive CPU Time
(ms)Application
32.950046290WhatsApp
3057040.04210Viber
64.9927065.613130WeChat
55.9476057.35490Imo

4.1.3 android.view.View.layout
TABLE 3: CPU Utilization of android.view.View.layout.
From the above table 3, it is observed that in case of View layout module WeChat CPU Utilization
is more as compared to WhatsApp, Viber and Imo.
4.1.4 android.view.ViewGroup.layout
TABLE 4: CPU Utilization of android.view.ViewGroup.layout.
From the above table 4, it is observed that in case of less load CPU Utilization of Wechat is more
as compared to WhatsApp, Viber and Imo whereas WhatsApp is taking the lesser time and in
case of more load Imo is taking the maximum time.
Hence it is concluded that Viber and WhatsApp are better in case of CPU Utilization factor
whereas WeChat and Imo are wasting the resource by more utilizing CPU.
4.2 Memory Utilization
The table data shows that all the applications before started occupy 75% of the total allocated
heap size.
ID Application
Heap
Size
(MB) Allocated(MB) Free(MB) %used #Objects
1956 WhatsApp 10.902 6.902 4 63.31% 79577
1845 Viber 10.992 6.992 4 63.61% 48814
914 WeChat 14.914 10.914 4 73.18% 69901
1283 Imo 6.536 3.921 2.614 60.00% 39218
TABLE 5: Memory Analysis after launching of the mentioned apps.
4.2.1 WhatsApp Memory Leak Suspects
Below are the major Problem Suspects and their Biggest instances found during WhatsApp
α =30α =3
Inclusive CPU
Time%
Inclusive CPU Time
(ms)
Inclusive CPU
Time%
Inclusive CPU Time
(ms)Application
4.8304.830WhatsApp
5.811013.570Viber
14.821208.91790WeChat
30.7261014.51390Imo
α =30α =3
Inclusive CPU
Time%
Inclusive CPU Time
(ms)
Inclusive CPU
Time%
Inclusive CPU Time
(ms)Application
4.8304.830WhatsApp
5.811013.570Viber
14.821208.91790WeChat
30.7261014.51390Imo

Launching:-
a) Problem Suspect 1
55,804 instances of "java.lang.String", loaded by "<system class loader>" occupy
4,834,912 (42.21%) bytes.
b) Problem Suspect 2
5,109 instances of "java.lang.Class", loaded by "<system class loader>" occupy
3,261,968 (28.48%) bytes.
Biggest instances measured in bytes are:
i.class com.whatsapp.App - 880,056 (7.68%)
ii.class com.whatsapp.fu - 246,520 (2.15%)
iii.class android.text.Html$HtmlParser - 126,616 (1.11%)
iv.class android.icu.impl.UCharacterProperty - 117,536 (1.03%)
c) Problem Suspect 3
14 instances of "java.lang.DexCache", loaded by "<system class loader>" occupy
1,704,672 (14.88%) bytes.
i.java.lang.DexCache - 547,136 (4.78%)
ii.java.lang.DexCache - 291,632 (2.55%)
iii.java.lang.DexCache - 253,176 (2.21%)
iv.java.lang.DexCache - 231,944 (2.02%)
FIGURE 1: Whats App Memory Leak Problem Suspects.
4.2.2 Viber Memory Leak Suspects
Below are the major Problem Suspects and their Biggest instances found during the Launch of
Viber:-
4,191,368 (32.72%) bytes.
1,865,432 (14.56%) bytes.

i. java.lang.DexCache - 550,168 (4.30%)
ii. java.lang.DexCache - 294,176 (2.30%)
iii. java.lang.DexCache - 253,176 (1.98%)
iv. java.lang.DexCache - 231,672 (1.81%)
v. java.lang.DexCache - 141,360 (1.10%)
1,619,808 (12.65%) bytes.
FIGURE 2: Viber Memory Leak Problem Suspects.
4.2.3 Wechat Leak Suspects
Wechat:-
4,493,696 (30.69%) bytes.
3,611,368 (24.67%) bytes.
Biggest instances are:
i. class com.tencent.mm.av.b- 1,423,864 (9.73%) bytes.
2,447,248 (16.71%) bytes.
i. java.lang.DexCache- 584,960 (4.00%)
ii. java.lang.DexCache- 547,088 (3.74%)
iii. java.lang.DexCache- 405,640 (2.77%)
iv. java.lang.DexCache- 253,888 (1.73%)

v. java.lang.DexCache- 231,744 (1.58%)
FIGURE 3: Wechat Memory Leak Problem Suspects.
4.2.4 Imo Leak Suspects
Wechat:-
4,078,880 (37.88%) bytes.
1,845,632 (17.14%) bytes.
Biggest instances are measured in bytes:
i. java.lang.DexCache (5.08%)
ii. java.lang.DexCache- 316,048 (2.94%)
iii. java.lang.DexCache- 252,992 (2.35%)
iv. java.lang.DexCache- 231,632 (2.15%)
v. java.lang.DexCache- 141,320 (1.31%)
1,534,144 (14.25%) bytes.
Biggest instances are:
i. class android.text.Html$HtmlParser- 126,616 (1.18%) bytes.
ii. class android.icu.impl.UCharacterProperty- 117,536 (1.09%) bytes.

FIGURE 4: Imo Memory Leak Problem Suspects.
4.2.5 Comparative Analysis
The following table shows that java.lang.String class is utilizing more memory in all the
applications. Hence it is concluded that java.lang.String class need to be efficiently used to
counter memory leak problem.
Class Name
Objects
#0
(WhatsA
pp)
Objects
#1
(Viber)
Objects
#2
(Wechat)
Objects
#3
(Imo)
Shallow
Heap #0
(Whats
App)
Shallow
Heap #1
(Viber)
Shallow
Heap #2
(Wechat)
Shallow
Heap #3
(Imo)
Retained
Heap -
Table
char[]7243154569602995411247220323675712411564036688564722032
byte[]259724832938208118713123152000272484013537921871312
java.lang.String7234354489602045402717362321307736144489612966486359416
int[]364630774235279116515281808824224533616808801651528
java.lang.String[]285515711869159813210001303144165776813235282020280
java.lang.Class54696255931650851273041212081920961170963388288
TABLE 6: Comparison Table of Memory Analysis.
4.3 UI Rendering Speed
The UI rendering speed comparison is done on the basis of the results obtained by three factors
i.e Measure, Layout & Draw. Profiling has been done to understand the rendering speed closely.
The table 7 shows that viber is taking lesser time time than rest 3 applications.
ApplicationNo. of ViewsMeasure(ms)Layout(ms)Draw(ms)Total(ms)
hatsApp3640000
Viber2040000
WeChat498201070100
Imo186107070150
TABLE 7: Comparison Table of of U I Rendering(Profile1).

ApplicationNo. Of ViewsMeasure(ms)Layout(ms)Draw(ms)Total(ms)
WhatsApp364020020
Viber204001010
WeChat4982030110160
Imo186104060110
WhatsApp3640000
Viber2040000
WeChat498204080140
Imo186105070130
WhatsApp364010010
Viber204010010
WeChat4982040100160
Imo18605070120
WhatsApp3640101020
Viber2040000
WeChat498204080140
Imo186104050100
TABLE 11: Comparison Table of of U I Rendering(Profile 5).
WhatsApp3640000
Viber204100010
WeChat498204080140
Imo186104050100
WhatsApp364020020
Viber2041001020
WeChat498204080140
Imo186104050100

International Journal of Computer Science and Security (IJCSS), Volume (13) : Issue (2) : 2019 33
ApplicationMeasure(ms)Layout(ms)Draw(ms)Total(ms)
WhatsApp08.57141.428610
Viber2.85711.42862.85717.1429
WeChat2034.28685.714140
Imo8.571447.14360115.71
TABLE 14: Comparison Table of Average Total of all the Rendering Profiles.
FIGURE 5: Graphical view of Grand Total of all the Profiles.
FIGURE 6: Graphical view of Table No. 17.
From the above Profile Tables and Figures , it has been shown that the rendering speed of the
Viber is less as compare to the rest of mentioned applications.
5. CONCLUSION
In this Paper, We have analysed the performance analysis of popular used apps like WhatApp,
WeChat, Imo and Viber. The performance of these apps have been evaluated in terms of CPU,
Memory & UI Rendering Speed. Experimental results yield that CPU utilization of WhatsApp and
Viber is better than others.
6. FUTURE SCOPE
The Study opens the window for android researchers to work on various Performance parameters
and it is recommended to analyze more modules in case of CPU Utilization. For Memory
assessment it is suggested to find out the shallow size results.
7. REFERENCES
[1] Lars Vogel. “Android application (performance and more) analysis tools”
http://www.vogella.com/tutorials/AndroidTools/article.html, July 5 , 2016[12-08-2016].

[2] Of Bytes and Battery. ”Of Bytes, Cycles and Battery Life”,
http://www.slideshare.net/mariozechner5/of-bytes-cycles-and-battery-life. Oct. 26,
2013[12-08-2016].
[3] Patrick Dubroy. ”Memory Analysis For Android Applications”, http://android-
developers.blogspot.in/2011/03/memory-analysis-for-android.html, March 24, 2011[14-08-
2016].
[4] Profile. “Android DDMS”, https://developer.android.com/studio/profile/ddms.html, March 24,
2011 [15-08-2016].
[5] MAT. ”Memory Analyzer (MAT)”, http://www.eclipse.org/mat/, April 15, 2016[19-08-2016].
[6] Command Line. “systrace”, https://developer.android.com/studio/profile/systrace-
commandline.html, April 15, 2016[19-08-2016].
[7] Roman Guy. “Android Performance Case Study”, http://www.curious-
creature.com/2012/12/01/android-performance-case-study/ , Dec. 1, 2012[20-08-2016].
[8] Profile. “Profile your layout with Hierarchy Viewer”,
https://developer.android.com/studio/profile/optimize-ui.html 20-08-2016
[9] The Statistics Portal. “Most popular global mobile messenger apps”,
http://www.statista.com/statistics/258749/most-popular-global-mobile-messenger-apps/ May
15, 2016[24-08-2016].
[10] Doug Sillars. ”CPU and CPU Performance”,
https://www.safaribooksonline.com/library/view/high-performance
android/9781491913994/ch06.html, Oct, 2015 [25-08-2016].
[11] Doug Sillars. ”Screen and UI Performance”,
https://www.safaribooksonline.com/library/view/high-performance-
android/9781491913994/ch04.html, Oct, 2015 [ 25-08-2016].
[12] Doug Sillars. “Memory Performance”, https://www.safaribooksonline.com/library/view/high-
performance-android/9781491913994/ch05.html, Oct, 2015 [25-08-2016].
[13] Android Studio User Guide. ”Test your app”,
https://developer.android.com/studio/test/index.html, April 4, 2016 [27-08-2016].
[14] Analyzing Java Memory. “Memory Management”,
http://www.dynatrace.com/en/javabook/analyzing-java-memory.html, April 4, 2016 [27-08-
2017].
[15] Inspect trace logs with Traceview. “Traceview”,
https://developer.android.com/studio/profile/traceview.html, April 5, 2016 [15-08-2016].
[16] MAT.”Memory Analyzer (MAT)”, http://www.eclipse.org/mat/, April 15, 2016 [19-08-2016].
[17] “How many app downloads are there per day for both the App Store and Google Play”,
https://www.quora.com/How-many-app-downloads-are-there-per-day-for-both-the-App-
Store-and-Google-Play, Nov, 2016[26-08-2016].
[18] Thiago Soares Fernandes, Erika Cota , Alvaro Freitas Moreira “Performance Evaluation of
Android Applications: a Case Study”, SBESC '14 Proceedings of the 2014 Brazilian

Symposium on Computing Systems Engineering, IEEE Computer Society Washington, DC,
USA, 978-1-4799-8559-3, 2014
[19] Liu Pu, “Performance Analysis Of Short Messages”, Intelligent Ubiquitous Computing and
Education, 2009 International Symposium on Intelligent Ubiquitous Computing and
Education, IEEE, 10859941, Chengdu, China, 2009.
[20] Anthony Gutierrez, Ronald G. Dreslinski, Thomas F. Wenisch, Trevor Mudge, Ali Saidi,
Chris Emmons, and Nigel Paver, “Full-system analysis and characterization of interactive
smartphone applications”, In Proceedings of the 2011 IEEE International Symposium on
Workload Characterization, IISWC ’11, pages 81–90, Washington, DC, USA, 2011.
[21] Dhinakaran Pandiyan, Shin-Ying Lee, and Carole-Jean Wu, “Performance, Energy
Characterizations and Architectural Implications of An Emerging Mobile Platform Benchmark
Suite – MobileBench”, IEEE International Symposium on Workload Characterization, IISWC
2013 - Portland, OR, United States, 2013.
[22] Rama Bhatia, Dr. Renu Bagoria, Dr. A. K. Mohapatra, 2016 “An analytical approach towards
CPU, Memory & UI Performance Assesment of Android Apps”, “International Journal Of
Control Theory and Application”,Volume : No.9 (2016) Issue No. :23 (2016) Pages : 247-
251.

Performance Comparison of Android Messengers

More Related Content

What's hot (15)

Similar to Performance Comparison of Android Messengers (20)

Recently uploaded (20)

Performance Comparison of Android Messengers