Access-Controlled Video/Voice over IP in
Hadoop
System with BPNN Intelligent Adaptation
Authors:
1
Speaker :童耀民 MA1G0222
2013.05.07
2
Outline
 INTRODUCTION
 VIDEO/VOICE OVER IP ON CLOUD COMPUTING
 REALIZATION OF CLOUD COMPUTING SYSTEM
 EMPIRICAL RESULTS AND DISCUSSIONS
 CONCLUSIONS
3
INTRODUCTION
 Real-Time Transport Protocol (RTP) [1], relying on
UDP, provides no loss recovery (unreliable).
 TCP protocol may cause long delay to get
obsolete data at receiver side.
 In the previous work [2], the phone based on
video/voice over IP (VVoIP) is established in
ARM-based embedded platform rather than x86
machines [3] used in the earlier project.
4
 The objective is to ring a mobile phone
call operated on P2P connection [4].
 In order to tackle three crucial problems:
headof- line blocking, handover
interruption, and non-real-time
transmission, we have adopted PR-SCTP
protocol [5] instead of TCP.
5
 However, the computation load for realtime video phone call is so big and has
caused ARM-based embedded
platform [6] no longer with power-saving.
 Moreover, VVoIP need to know their
respective IP address between two
mobile devices before P2P connection.
6
 In order to resolve the problems as mentioned above,
this study is to realize a real-time video/voice over IP
(VVoIP) applications on Hadoop cloud computing [7]
denoted CLCIHU.
 Next based on client/server architecture users employ
web interface to achieve VVoIP connection and the
phone call adopts an easy TCP-based RMTP connection
instead of the complicated PR-SCTP because of web
interface not supported by PR-SCTP.
7
 多媒體串流一般是透過傳輸層協定來傳遞，
像是TCP 或UDP，跟UDP 協定做比較，TCP
協定雖然可以保證資料的可靠性，但為了確
保資料的完整性，所以增加了整體傳輸的時
間。IETF(Internet Engineering Task Force)
在2000 年制定了一個新的傳輸層的通訊協定
稱做SCTP (Stream Control Transmission
Protocol)，跟TCP 協定一樣具有可靠的特性，
但其又具有多宿主(multi-homing)的特性，
SCTP改善了關連(association)的可靠度，當
主要的傳輸路徑發生錯誤時可將資料切換到
其它的路徑上傳遞。SCTP 在2004 年提出了
8
 In terms of the quality of services (QoS), we have to
tackle some of crucial problems about jitter, loss,
latency, and throughput so as to maintain the smooth
video/voice streaming over internet.
 We employed a back-propagation neural network
(BPNN) [12] to tune parameters appropriately for
adapting seamless handoff and analyzing network
traffic over time.

VIDEO/VOICE OVER IP ON CLOUD
9
COMPUTING
 A number of remarkable cloud computing
related firm, as we know, are indicated as
follows:
 (a) public cloud like Google App Engine,
Amazon Web Services, and Microsoft Azure;
 (b) private cloud like Microsoft MCloud, IBM Blue
Cloud, and SalesForce.com;
 (c) open cloud: Open Nebula, Eucalyptus,
Apache Hadoop, and NCDM Sector/Sphere;
 (d) hybrid cloud like IBM Blue Cloud.
10
 Two famous companies related to IP phone
on cloud computing are voice over IP in a
Cloud, IIS 2009 [13] and VoIP in cloud
computing, Skpye in 2010 [14].
 It has been set up in Hadoop cloud
computing server [15][16] so that web
interface is applicable to run and browse
video phone call between cloud computing
server and thin clients as shown in Fig. 1.
11
REALIZATION OF CLOUD COMPUTING
12
SYSTEM
 A. Deploying Hadoop cloud computin
 Once a Hadoop cloud computing has
been established in server site, we have
to test a number of seamless handoffs
between clients and a Hadoop cloud
computing system as shown in Fig. 2, 3,
and 4.
13

14B. Establishing thin client and installing access
control
 This study has chosen QT/Embedded framework
instead of GTK series, in such a way that achieves
GUI interface functions.
 Afterward QT/Embedded gets through the kernel
driver to activate graphic function.
15
 Cloud computing here can perform
rapid fingerprint identification and facial
recognition in order to fulfill the access
control system [21].
16C. Implementing VVoIP at Hadoop with web
interface
 (1) According to an open source IHU (a voice over
IP application program), we have established a
video/voice over IP application on Linux system to
develop VVoIP over two PCs as shown in Fig. 6
17(2) Activating a database in a cloud computing
server to record the necessary information like
account, IP address and so on.
 (3) Constructing the web services in order to connect
a database in Hadoop cloud computing server for
implementing the VVoIP applications, user account
and login, as shown in Figs. 7 and 8.
 (4) Transplanting the video/voice over IP application
into cloud computing server incorporation with web
interface and database.
18
19
 D. Intelligent adaptation for VVoIP on
cloud computing
 In terms of quality of services (QoS)
implemented in the video phone call
over IP, we have to tackle some of
crucial problems about jitter, loss,
latency, and throughput so as to
maintain the smooth video/voice
streaming over internet.
20
EMPIRICAL RESULTS AND
DISCUSSIONS
21
22
23
24
25
CONCLUSIONS
 In this study we have moved the VVoIP
applications from user devices to Hadoop cloud
computing system with access control to
implement rapid facial/fingerprint identifications.
 User does not need to use a real IP and have a
friendly web interface.
 This scheme reduces computation load and
power consumption dramatically at user devices
In addition, we employed a BPNN to tune
parameters appropriately for adapting seamless
handoff and analyzing network traffic over time.