A Novel P2P VoD Streaming Technique Integrating Localization and
Congestion Awareness Strategies
This material is presented to ensure timely dissemination of scholarly and
technical work. Copyright and all rights therein are retained by authors or
by other copyright holders. All persons copying this information are
expected to adhere to the terms and constraints invoked by each author's
copyright. In most cases, these works may not be reposted without the
explicit permission of the copyright holder.
Citation:
Mostafa M. Fouda, Zubair Md. Fadlullah, Mohsen Guizani, and Nei Kato, “A
Novel P2P VoD Streaming Technique Integrating Localization and
Congestion Awareness Strategies,” Springer Mobile Networks and
Applications (MONET), vol. 17, no. 5, pp. 594-603, October 2012.
URL:
http://www.springerlink.com/content/463u11672g3v0471/
A Novel P2P VoD Streaming Technique Integrating
Localization and Congestion Awareness Strategies
Mostafa M. Fouda · Zubair Md. Fadlullah · Mohsen Guizani · Nei Kato
Abstract The concept of the “future Internet” has
evolved amongst researchers recently to relieve the tremen-
dous pressure on the current Internet infrastructure
to support the heterogeneous networking technologies,
mobile devices, increased population of u se r s , and also
the high user requirements for r e al-ti me services and ap-
plications. Peer-to-Peer (P2P) Video on Demand (VoD)
streaming technologies are expected to be a key tech-
nology in the future Internet. Because the existing P2P
streaming techn iq ue s are attributed to a number of
shortcomings, P2P VoD schemes need to be adequately
redesigned for the future Int er n et . In this paper, we
propose a scheme to effe ct ively provide VoD by using
P2P-based mesh overlay networks that may be suitable
for the future Internet. Our scheme selects the most
appropriate peers by exploiting domain-based localiza-
tion and congestion awareness strategies. Through sim-
ulations, our proposed scheme is demonstrated to have
scalability and capability of reducing the startup de-
Part of the work has been presented in Chinacom’10 [1].
M. M. Fouda
Graduate School of Inform ati on Sciences (GSIS),
Tohoku University, Japan
Tel.: +81-22-7956161
Fax: +81-22-7956161
E-mail: mfouda@it.ecei.tohoku.ac.jp
Z. M. Fadlullah
Graduate School of Inform ati on Sciences (GSIS),
Tohoku University, Japan
E-mail: zubair@it.ecei.tohoku.ac.jp
M. Guizani
Qatar University, Qatar
E-mail: mguizani@ieee.org
N. Kato
Graduate School of Inform ati on Sciences (GSIS),
Tohoku University, Japan
E-mail: kato@it. ecei.t ohok u. ac.j p
lay and total link cost, while maintaining high play-
back rate. The results are encouraging and show the
importance of redesigning P2P VoD services in futur e
Internet.
Keywords Peer-to-Peer · Video-on-Demand stream-
ing · Future Internet
1 Introduction
Over the last decade, the capacities of the Internet in
both access and core networks have vastly improved.
According to the estimation of Gartner, the worldwide
broadband installations is expected to reach an over-
whelming figure of 580 million households in 2013, i.e.,
the popularity rate of broadband subscribers will b e
around 25% [2]. As per a report from Cisco, the growth
of the Internet in 2015 will be four times larger than
in 2010 [3]. The fast deployment of broadband access
networks enables massive data del ivery at a fast pace.
Instead of being restricted with delivery of only text
and images, broadband Internet brings diverse audio
and video streaming applicati ons .
Also, recently, a huge variety of s ervices have evolved
in order to sati s fy the ever-increasing demands for s har -
ing information. However, because the existing Internet
architecture was designed four decades ago, in the n e ar
future it may falter under the tremendous pressure cre-
ated by the increasing users’ requirements in terms of
Quality of Service (QoS). In addition, the current Inter-
net architecture also lacks a number of featur es like reli-
ability, resiliency, s ec ur i ty, mobility, context-awareness,
and so forth. To address this issu e, two projects were
initiated by the National Science Foun d ation (NSF) in
the United States called Global Environment for Net-
work Innovations (GENI) [4] and Future InterNet De-
sign (FIND) [5]. Both the GENI and FIND projects
2 Mostafa M. Fouda et al.
focused on addressing the need for cross-layer Inter-
net design, network virtualization, dynamic switching
of opt ic al c ir c ui ts , s e r vi ce di sc overy and composition,
service management, traffic and routing engineerin g,
along with many other directions. With the similar ob-
jectives in mind, the European Commission also ini-
tiated its own project termed as the Future Internet
Research and Experimentation (FIRE) [6] and that is
within the 7
th
Framework Program (FP7). Indeed, the
proposals in FIRE indicated the clear need to carry out
early experimentation and testing over large-scale envi-
ronments rather than solely relying on theoretical def-
initions of next generation Internet architectures, pro-
tocols, and services. The AKARI project in Japan was
initiated by the National Institute of Information and
Commu ni cati ons Technology (NICT) [7] and it stressed
upon gen er ati ng ideas, technologies, and a NeW Gener-
ation Network (NWGN) architecture by the year 2015.
The above-mentioned projects and re s ear ch end eavors
have identified one of the ser ious shortcomings of the
current Internet that consists in dealing with traffic vol-
ume. The network operators are required to offload traf-
fic at optimal points of the current Inte r n et framework.
Recent ye ars have seen a rapid growth of video traf-
fic over t he Internet. According to [3], Internet video is
now 40% of all consumer Internet traffic, and will ac-
count for 62% of all consumer Internet traffic in 2015.
To this end, a major goal of the future Internet projects
is to envision improved means of handling data traf-
fic, substantial portion of which is currently generated
by video streaming. Video streaming can be catego-
rized into two streaming techniques, n amely live video
streaming and Video on-Demand (VoD) streaming.
Live video streaming is a broadcast of a li ve tele-
vision channel or any othe r video content to multiple
users at the same time. The traditional way to do this
is to send the same vid eo chunks to each client sepa-
rately which r eq ui r es huge bandwidth. In addition, th is
technique is not scalable. Another solution to do this is
by using Internet Protocol (IP) multicast technology. In
IP multicast, the server sends the video chunks to mul -
tiple assigned clients with only one traffic flow. Using
this technology should avoid having a bottle-neck near
the sour ce since only one traffic is being transferred.
The main problem with IP multicast is that not all the
routers in th e curr ent Internet infrastru ct ur e support
such kind of technology which makes it a rather re-
stricted technology. To this end, a cheap e r and reliable
technology is needed to en able live video streaming over
the Internet in an efficient manner.
On the other hand, Vid eo on-Demand (VoD) stream-
ing is more popular for Internet users since it provides
user-interactive facilities (e.g., pause, random jumps,
rewind, forward, slow motion, and so forth). One of
the popular examples of VoD streaming is “YouTube”
where Internet users can watch any part of a video
anytime. With these more advanced features of VoD
over live video streaming, multicasting cannot present
itself as an adequate solution. As we mentioned earlier,
sending separate chunks to individual costumers is also
expensive. So, it is re commend ed to design a cheaper
and reliable tech nology, which needs to be carefully de-
signed to achieve low cost, decent Q uali ty of Experience
(QoE), and high scalability.
Three different ways do exist for sendin g such kind
of tr affic. The first technique is the traditional client-
server paradigm whereby a single server sends the data
to consumers. The second way consists in the Con-
tent Distribution Network (CDN), which is designed to
avoid bottleneck near that server by having duplicated
servers in different regions. The third way is the Peer -
to-Peer (P2P) overlay networks where the resource s of
the client s are utilized for uploading the content with-
out the need for a server. It should be noted th at hybrid
technologies also exist. However, the cheapest technol-
ogy is based on Peer-to-Peer (P2P), since it is com-
pletely based on clients’ resources .
P2P applications are employed for sharin g files and
streaming live/on-d emand multimedia contents. These
P2P applications involve a huge volume of informa-
tion exchanged along both uplink and downlink direc-
tions amongst users/peers. Compared to the more com-
monly used client/server-based applications evolved in
the traditional Internet, P2P programs often select one
or more suitable candidates from a pool of peers, which
offer the same services or resourc es . This unique fea-
ture of P2P applications leads to incr e ase d reliability
and resiliency to a single point of f ailu r e. As a conse-
quence, P2P applications are expected to be a vital part
of future Internet.
The aforementioned points make the design and de-
ployment of a P2P-VoD syste m in the future Internet
rather difficult due to compl ex features of VoD ser-
vices. Therefore, the research work on developing P2P-
VoD architectures to fit the future Internet framework
is a timely and important one. In order to effectively
overcome this problem, we envision a novel P2P VoD
scheme in this pape r .
The remainder of this paper is structured as fol-
lows. Section II su r veys several future Internet initia-
tives and also presents related research work on devel-
oping P2P VoD schemes. An overview of real time VoD
P2P streaming systems is provided in Section III. Next,
in Section IV, our envisioned P2P-based VoD system is
presented. The simulation results are provided in Sec-
tion V to demonstrate the effectiveness of the proposed
A Novel P2P VoD Streaming Technique Integrating Localization and Congestion Awareness Strategies 3
scheme. Concluding remarks are presented in in Sec-
tion VI.
2 Related Work
There is a number of recent initiatives in order to find
the specific requirements of the future Internet. This
section first de s cr i bes several future Internet initiatives
or projects carried out in the recent past. Following this,
a number of P2P VoD schemes introduced by recent re-
searchers that may be incorporated with the upcoming
future Internet are surveyed.
2.1 Recently Conducted Future Internet Projects
The Internet has come a long way since its conception in
the 1970s. The very concept that led to t he fascinating
breakthrough in Internet communications has become
rather obsolete as it was not originally designed to ad-
dress the current user-requirements for various services
over heterogeneous networks. In fact, it is surprising to
many that the Internet has survived so many decades of
pressure from a wide variety and ever-increasing num-
ber of wired and wireless mobile users and applications,
state-of-the art business models, latest networking de-
vices and equipment (e.g., multiple wireless interfaces
that enable users to connect to different networks at
the same time [8]), and so on.
Therefore, these limitations of the Internet are taken
into serious consideration by leading researcher s which
have prompted them to resort to a concept of providing
a futurist ic Internet framework.
Currently, a number of research projects are being
carried out to develop the future Internet. One of these
notable projects for dealing with the challenges of to-
day’s Internet and envisioning future Internet networ k s
is referred to as FIND [5]. The FIND project was car-
ried out by NSF. FIND stresses on the developing net-
work architecture, security, advanced wireless and op-
tical properties, economical principles, and in general,
the means to effectively construct a global network in
fifteen years from now. Another project called GENI [4]
program was initiated to address the absence of secu-
rity, reliability, evolvability, and manageability of the
current Internet. In Europe, there were also similar
projects. For example, the FIREworks (Future Inter-
net Research and Experimentation - Str ate gy Works)
project [9] and the EIFFEL (Evolved Internet Future
for European Leadership [10]) Support Action (SA) for
the 7
th
Framework Program (FP7) are prominent fu-
ture Internet initiatives taken by European countries.
Although these different projects have the same end
goal (i.e., facilitating future Internet) in their focus,
their means are not necessarily similar. In f act, the
contrast in design goals is als o present in the various
Internet Engineering Task Force (IETF) work groups.
For example , one IETF group envisioned Low Extr a
Delay BAckground Transport (LEDBAT) protocol [11]
for minimizing the additional delay caused by sophisti-
cated networking applications while another one intro-
duced the Diffserv (Differentiated serv ic es ) Code Point
for bulk traffic to identify, mark, and manage congestion
events arising from P2P traffic under Diffserv frame-
work. Furthermore, to manage the P2P t r affic volume
within an acceptable level at the network operators’ ex-
pensive links, two IETF groups were formulated, namely
DECoupled Application Data Enroute (DECADE) [12]
and Application Layer Traffic Optimization (ALTO) [13]
working groups. DECADE identifies the problem with
current Inter ne t caches (e.g., P2P and web caches),
which attempt to provide adequate storage capabilities
within the network for reliable access of resources. How-
ever, DECADE is not able to explicitly support individ-
ual P2P applicati on protocols or user access to the con-
tent providers’ caches. Th e latter attempts at providing
a simple means for conveying network information to
the concerned P2P applications. This assists these P2P
applications to mitigate overhead of measuring topol-
ogy information. Furthermore, the work in [14] sug-
gested, instead of random initial peer selec tion , adop-
tion of traffic localization for P2P applications by in-
troducing ALTO service.
2.2 Related wor k on P2P VoD schemes
To me et the requirements of future Internet, a number
of research works have emerged recently in literature
that have attempted to combine P2P strategies with
the server-client streaming framework. A note-worthy
example is (BASS) [15], which comprises an external
media server and a modified BitTorrent [16] protocol.
The clients in BASS are not able to download contents
prior to the current playback time. Instead, these clients
receive chunks from the media server sequentially. Also,
they do not need to obtain the already/being down-
loaded chunks by the BitTorrent. However, the BASS
framework lacks scalability when the sy st em serves a
large number of peers .
To deal with the above pr oble m, no external media
server was taken into account in the BitTorrent Stream-
ing (BiToS) framewor k [17]. Rather than adopting the
“rarest first” strategy, BiToS use s a selection mecha-
nism, which ignores missing chunks (i. e., the chunks un-
able to meet their playback time). Due to these chunks
4 Mostafa M. Fouda et al.
being missed during the download process, the video
playback is disrupted.
The work in [18] focuses on providing small startup
delays in P2P-VoD systems by integrating network cod-
ing with segment scheduling. Thus, it achieves a high
utilization of the available resources. However, this ap-
proach does not consider practically deployable P2P
VoD environments. In order to overcome this problem,
it is important to design a s cheduling mechanism capa-
ble of instructing the involved peers to assist one an-
other in streaming the VoD contents in a real-time man-
ner. To this end, the auth ors proposed a VoD scheme
in an earlier work [19] over P2P mesh-based overlay
networks with user-scalability features.
3 Overview of conventional P2P VoD streaming
Along with advances in broadband technologies, multi-
media communications such as interactive on-d emand
video streaming services have become popular Inter-
net applications in recent time (e.g., YouTub e) . In the
traditional client/server setting, videos are delivered to
the users through a centralized entity/server. Two of
the main drawbacks of such a VoD scheme are the
lack of scalability (i.e., support of a l imit ed number
of subscribers) and the limited streaming rate at the
end-users which is due to the limited bandwidth of the
server. P2P-based streaming has recently emerged as an
alternative to traditional server-based streaming (e.g.,
Youtube) for VoD provisioning. It already emerged as
a promising technology for the future Internet target-
ing global users connected to heterogeneous networks.
From the perspectives of a broadcaster/content provider,
the P2P approach provides an added incentive as it al-
lows the serv ic in g of a lar ge audience (i.e., many peers)
without the need for investment in additional resour c es .
On th e other hand, a user also experiences improved de-
livery rate of the multimedia content while he/she can
also upload his/her own content to other peers. The
implication of use of P2P VoD in future Internet is im-
mense, as it tolerates only minimal change to the ex-
isting infr ast r uc tu r e. In addition, P2P VoD approaches
may also overcome bandwidth /pr ocessing load bottle-
necks, reduce st art up and end-to-end delays, and also
improve the playback rate of the video. The existing
P2P VoD systems can be broadly classified as either
tree-based or mesh-based ones. In the tree-based sys-
tems, every node receives data from a source/parent
node acc ord in g to a tree-like structure. This intrinsic
design of tree -b ase d P2P VoD systems leads to a sig-
nificant problem in future Internet dynamics whereby
the nodes are expected to fr eq ue ntly change resulting
in frequent re-constr uc tion of the trees. This will ren-
der the children nodes to fail to obtain the streaming
feed until the tree is re-constructed. It is worth noting
that Application Layer Multicast (ALM), which aims at
replacing IP multicasting for content delivery, also s uf -
fers from this shortcoming [20,21]. ALM nodes, depend-
ing on the streaming application, constr uc t a multicast
tree, through which the stream is delivered. However, if
a node leaves the tree, it is unable to furthe r deliver the
stream to its descendant no de s till the tree is recovered.
In case of the mesh-base d overlay network, a peer,
which has j us t joined the overlay, contacts the tracker
(e.g., a directory server) to receive a list of a number
of active peers. After receiving the same, this new peer
immediately starts to requ es t the video chunks from
the peers in the neighbor-list. Also, note that a peer
is connected with a small subset of active peers at any
time, and it is permitted to exchange video chunks and
control messages only with them. The overall architec-
ture of the mesh-based P2P VoD is illustrate d in Fig. 1.
Also, it is worth-noting that the mesh-based P2P VoD
networks are more resilient to node failures compared
with thei r tree-based c ounterpart.
Another weak assumption made by most of the ex-
isting P2P-based VoD systems is as follows. A new user
keeps watching the video stream until (i) the user leaves
the mesh overlay, or (ii) the streaming session fails.
This assumption does not take into account the user-
interaction features ( e. g., pausing, replaying, fast for-
warding, and so forth) in order to simplify the system
design. In practice, instead of continuous ly playing the
video, the VoD subscribers tend to often jump to a more
interesting section of the video either because they do
not feel interested to watch that segment or do not have
enough time to watch the entire video. The frequent
interactions of the user s in the P2P VoD applications
present a significant challenge on the continuity of the
video p layback in terms of the “seeking delay”, which
refers to the time required since the re qu es t for a video
segment till the segment becomes available. In orde r
to ensure an almost zero seek in g delay to watch the
video without disruption, each video s egment requir es
to be pre-fetched by a peer prior to the playback of the
segment. This enhances t he playback continuity. Most
conventional P2P VoD systems, however, as mentioned
earlier, based on their assumption that the us er s watch
the vid eo sequentially without any interactive “seek”,
perform the “pre-fetching” of video segments in a se-
quential manner. An alternative to this approach is to
adopt a mechanism that pre-fe tches random video seg-
ments. In this vein, we propose a pre-fetching mecha-
nism for obtaining appropriate video segments a priori
