ROBUST MULTIMEDIA TRANSMISSION OVER WIRELESS
AND MOBILE NETWORKS
Nikolaos Papaoulakis, Charalampos Z. Patrikakis, Mario S. Nunes and Rui S. Cruz
Keywords: Wireless Networks, Multimedia, Transmission, Optimisation.
Abstract: In this paper, an analysis of the different techniques for supporting robust multimedia transmission over
wireless media is given. The analysis includes Radio Resource Management techniques on the Physical
layer, transmission techniques on the Network (IP) layer, optimisation techniques on the Transport layer and
techniques focusing on the Application layer. Also there is a report on the selection of the most efficient
solutions and the way these can be combined in an integrated and cross layer optimisation solution. The
paper has been prepared following the results of the research performed in the context of the ICT project
my-eDirector 2012 in order to support the robust transmission of live streaming services for the coverage of
athletic events for large numbers of heterogeneous networked users.
1 INTRODUCTION
Robust real-time IP multimedia streaming support
over mobile networks is a very intriguing task.
Compared to fixed networks, multimedia
transmission over wireless channels is a far more
challenging research topic, due to the uncertainness
and limitations of the air interface such as the high
Bit Error Rates (BER), the limited throughput, the
heterogeneity of network nodes and the
corresponding impact that these have on the
compressed media domain. Additionally, another
issue is the guarantee of Quality of Service (QoS),
during network congestion and user mobility
(handover etc). In existing implementations, each
network layer tries to address these challenges
separately, through the deployment of its own
adaptation and protection mechanisms. However,
this strategy cannot not always guarantee an optimal
overall performance.
In this paper, an analysis of the different
techniques for supporting robust multimedia
transmission over wireless media is given, and a
concluding report on solutions’ efficiency and the
way each one addresses the issue of robust
multimedia transmission over wireless and mobile
networks is presented. The analysis has been
performed in the context of selecting the most
appropriate mechanism (or combination of
mechanisms for the implementation of a platform for
supporting real time streaming services for large
numbers of users over heterogeneous networks,
focussing on the coverage of live athletic events
(Patrikakis et al., 2010). This is the main objective
of the ICT My-e-Director 2012 project, and the
results presented in this paper derive from the
research work performed in the scope of this project.
2 ROBUST MULTIMEDIA
TRANSMISSION
2.1 Load Balance in IEEE802.11
The main problem that wireless networks have to
face is the non-uniform traffic distribution along the
base stations of the network. The radio resource
management techniques that currently interact with
the Medium Access Control (MAC) layer
mechanism of wireless networks may be seen as
added value techniques for increasing the network
performance and the supported QoS.
Dynamic load balancing and congestion control
aim to uniformly distribute idle or active users along
the dominance or serving areas of each base station.
The left side of Figure 1 represents a non-uniform
idle user distribution across the sectors of a base
station. The right side of the same figure represents
the ideal scenario, i.e., the number of users per
sector follows a far better distribution.
142
Papaoulakis N., Z. Patrikakis C., S. Nunes M. and S. Cruz R. (2010).
ROBUST MULTIMEDIA TRANSMISSION OVER WIRELESS AND MOBILE NETWORKS.
In Proceedings of the International Conference on Wireless Information Networks and Systems, pages 142-147
DOI: 10.5220/0002953201420147
Copyright
c
SciTePress
Figure 1: Load balancing problem.
In IEEE 802.11 standards (also known as WiFi
standards) for Wireless LANs (WLANs), there is no
inherent mechanism for QoS provision, as these
networks have been designed driven by the need for
low cost deployment and use of conventional IP
services like WEB browsing, FTP and email.
As WLANs have become a prevailing
technology for broadband mobile services, WiFi
devices have also been expanded to support a much
wider range of applications, such as VoIP
(VoWLAN) and multimedia, including WiFi
adapters/bridges that are able to deliver real time
media streams to the end user. The real time media
streaming, which is one of the most QoS demanding
applications, introduces the imperative need for
Radio Resource Management (RRM). The
requirement for guaranteed QoS brought to the
foreground the need for enhancing the WLAN
Medium Access Control (MAC) protocols with
appropriate QoS support mechanisms, and for this
reason, the IEEE 802.11e standard enhanced the
Distributed Coordinated Function (DCF) and the
Point Coordination Function (PCF) by a new
coordination function, the Hybrid Coordination
Function (HCF). Since IEEE 802.11e describes a
packet scheduling mechanism on the MAC Layer
that is compatible with existing IEEE 802.11 WLAN
standards, existing Access Points (APs) can be
upgraded to comply with IEEE 802.11e through a
relatively simple firmware upgrade.
Furthermore, in IEEE 802.11 WLANs, there is
the need for distributed RRM techniques that
perform load balancing of the traffic among all the
APs of the infrastructure network, for a more
efficient use of the scarce radio resources. Load
balancing at this level is able to provide a cross layer
QoS provision in all IP services even in multi-
vendor and high mobility environments. In these
networks, the Mobile Station (MS) has the
functionality to select an AP, based on specific
criteria (mainly focussing on the selection of the AP
with the strongest receive signal). A key challenge is
how to achieve overall load balancing in the network
during the AP reselection procedure for the optimum
utilization of network resources.
2.2 Transmission Modes on the IP
Layer
Apart from the physical medium related techniques
for efficient management of available resources,
alternative techniques to either wired or wireless
networks may also be considered to ease the load in
situations with high traffic demand. Such techniques
include the selective deployment of unicast,
multicast and broadcast transmission modes. Though
the use of unicast transmission is dominant in wired
networks, the shared media in wireless and mobile
access (air) make the use of multicast and broadcast
more appealing and far more effective than in wired
networks. In the following we will evaluate the use
of these techniques over different Radio Access
Technologies.
Unicast transmission mode is the most common
transmission mode for IP services, since it is linked
to widely used applications such as web page access,
e-mail, FTP based file transfer or telnet, and is also
supported in all radio access technologies, including
DVB-H IP datacast.
Multicast transmission incorporates the
capability of easy scaling transmission to a large
receiver population without requiring prior
knowledge of the IP addresses of the receivers or the
number of the receivers. Even though multicast
relies on the use of UDP protocol, not able to
guarantee reliable and error free transmission,
reliable multicast protocols such as Pragmatic
General Multicast (PGM) [RFC3208] have been
developed to add loss detection and retransmission
on top of IP Multicast. Multicast is supported by all
radio access technologies, like the Multimedia
Broadcast Multicast Service (MBMS) that can be
offered via existing UMTS cellular networks.
Multicast transmission over wireless networks has
the obvious advantage of optimizing radio resource
utilisation, but also has some limitations, like the
heavy packet loss that occurs during AP reselection
due to user mobility, as is the case with the very
simplistic approach of IEEE 802.11 networks.
Broadcast transmission makes the most
efficient use of network resources. However, this
transmission mode can only be applied in a local
subnet, as the routers by default do not forward
broadcast packets, to prevent networks floods. In
wireless access networks, broadcast transmission
can be adopted over single hop networks like WiFi
per AP, WiMAX per base station, or on DVB-H, but
ROBUST MULTIMEDIA TRANSMISSION OVER WIRELESS AND MOBILE NETWORKS
143
the only visible use of the broadcast transmission at
this point is essentially over DVB networks due to
their unidirectional nature.
2.3 Optimisation Techniques on the
Transport Layer
Media streaming on IP networks is challenging,
especially when end-to-end connections extend over
wireless networks with many factors such as
interferences, multipath fading, user mobility and
other general conditions that may cause errors that
result in frame losses. Therefore, some optimisation
techniques can be applied at the transport layer for
QoS adaptation and robust media transmission, as
are the cases of using Datagram Congestion Control
Protocol (DCCP) (Kohler et al., 2006), for
connectionless oriented services, TCP-Friendly Rate
Control (TFRC) (Floyd et al., 2006), for connection
oriented services or Stream Control Transmission
Protocol (SCTP) (Stewart et al., 2006), as well as
fine tuning techniques for TCP and UDP protocols.
Datagram Congestion Control Protocol
(DCCP) is a recently standardized protocol filling
the gap between TCP and UDP protocols. Unlike
TCP, it does not support reliable data delivery and
unlike UDP, it provides a TCP-friendly congestion
control mechanism in order to behave in a fair
manner with other TCP flows. DCCP includes
multiple congestion control algorithms, through its
Congestion Control ID (CCID), which can be
selected in regards to the user QoS requirements.
The rationale behind the use of DCCP is its inherent
capability for better handling of the multimedia
traffic and provision of a degree of transmission
control for real time.
TCP-Friendly Rate Control Protocol has been
designed to provide an equation-based congestion
control protocol using UDP for transport together
with optimal throughput estimations performed at
application level. The main goal here is to be able to
support optimized multimedia flow based on unicast
transmission over best-effort Internet environment
(Floyd et al., 2006). In wireless networks however,
TFRC still led to poor performance of throughput
(Zhou et al., 2007).
Stream Control Transmission Protocol
(SCTP) can be used as the transport protocol for
media streaming services, where monitoring and
detection of data loss and delay is required
(Rajamani et al., 2002), as it is a reliable transport
protocol operating on top of the potentially
unreliable connectionless IP packet service (Stewart
et al., 2000). Its design includes inherent support for
congestion avoidance, as well as the corresponding
mechanisms for resisting to flooding and
masquerade attacks. For such applications, the SCTP
path/session failure detection mechanisms will
actively monitor the connectivity of the session.
SCTP distinguishes different streams of messages
within one SCTP association, where only the
sequence of messages needs to be maintained per
stream. This approach helps in avoiding head-of-line
blocking problems between independent streams of
messages. However, use of SCTP has some
disadvantages related to flow control, selective
acknowledgement, congestion control and
multiservice support.
Media streaming over TCP is not effective
enough for streaming applications due to the
window based congestion control that doesn’t
provide instant rate adaptation and by the use of byte
stream and single connection by the peer ends. Such
characteristics can degrade TCP performance as the
TCP sender is not in position to distinguish the
origin of packet losses is due to wireless medium
degradation or to congestion in the network,
resulting in unnecessary congestion control actions;
however link errors in radio networks can be faced
up by transmission power regulation, code
redundancy (FEC) or retransmissions (ARQ).
Media Streaming over UDP is lighter and faster
than over TCP with better results in throughput.
However, due to the unreliable nature of the
protocol, the protection mechanisms commonly used
introduce extra overhead. A technique that may
overcome such problems is efficient header
compression that brings the additional advantage of
significant reductions in bandwidth requirements.
An alternative is the use of UDP Lite, as this
protocol exploits the redundancy that lies in IP layer
information (from which the UDP length may be
yielded) by replacing the “length” field of UDP with
a “coverage” field and therefore dividing packets
into sensitive and insensitive parts.
2.4 Optimisation Techniques on the
Application Layer
There are several adaptation techniques that can be
applied to support quality aware multimedia content
transmission (Santos et al., 2009). In order to
increase the efficiency of the adaptation techniques,
information related to the networking conditions, the
client playback environment as well as the specific
architecture that will be called to apply these
adaptation techniques. In particular, the information
WINSYS 2010 - International Conference on Wireless Information Networks and Systems
144
that needs to be known prior to the application of the
adaptation consists typically of:
Characteristics of the client device (size of
display, colour depth, buffer size and the
hardware and software types).
Characteristics related to the content (content
buffer size, minimum streaming bitrate,
compression formats and hardware and
software requirements).
Characteristics of the connected networks
(mobile or wired, network technology,
bandwidth, jitter, packet loss, delays and
channel variations).
For the adaptation point location and its architecture,
three scenarios can be considered. In the first, the
client device sends messages periodically to the
server, with information about the variations of the
channel characteristics. Based on that information,
the server decides on applying the adaptation
techniques. In the second the server sends the
original content to the client without adaptation. The
client receives the content and adapts it according to
its characteristics. In the third, the client sends to a
Proxy messages containing information of both its
own characteristics and those of the network to
which it is connected. The Proxy intercepts the
original content from the server, adapts it according
to the client conditions and forwards it to the client.
The adaptation techniques can be categorized in
three classes (Santos et al., 2009):
Format conversion. This procedure transcodes
the original content to another format (e.g.,
MPEG-4 to MPEG-2), compatible with the
ones supported by the client device.
Selection/Reduction. These techniques are a
trade-off between the content resources and
the characteristics of the network, spanning
from reduction on the number of frames per
second to reduction of the resolution.
Substitution. This class of adaptation technique
proposes the replacement of certain elements
of the content for other types of elements with
less impact in the bandwidth. As an example,
a live stream video can be replaced by a
slideshow, containing the key-frames of the
original video.
For the architecture of the Adaptation System,
several approaches are possible, being the most
relevant on RTSP and RTP transport, the
InfoPyramid
(Mohan et al., 1999), the Dual Point
(Hutter et al., 2005) and the Context Merging (Hutter et
al., 2005) architectures and, based on HTTP
transport, the emerging HTTP Adaptive Streaming
architecture (Patrikakis et al., 2009), (Cruz et al.,
2009), (Zambelli, 2009.
The InfoPyramid architecture (Figure 2) is
typically located in a Proxy and is modular.
Multimedia contents requested by clients are stored
on a Content Source module from where a Content
Analysis module extracts adequate information for
the adaptation process running on the InfoPyramid
module where the adaptation techniques are applied.
The selection of the adaptation techniques in the
Customization/Selection module is based on the
content and on information about the client device
capabilities (information gathered by the Client
Capabilities module).
Figure 2: InfoPyramid adaptation (Mohan et al., 1999).
The Customization/Selection module will then
select the adapted content (from the InfoPyramid
module) that maximises the quality of the content
(for that client) and delivers it.
The Dual Point architecture is also located in a
Proxy. In this architecture, clients requesting the
same content to the server provide context
information (device and network characteristics) to
an Adaptation Node. This node compares all the
context information in its knowledge base to the
ones of these clients to verify the suitability of the
content sizes that satisfy their requests, delivering
them if adequate or requesting a new minimum size
content able to satisfy all the clients in the same
conditions that are connected to it. This new
minimum size content is adapted to be delivered to
individually each client. This architecture only uses
resolution reduction technique to the content but
applied in two locations: in the Server, that adapts
the content to the minimum size, and in the
Adaptation Node that adapts the size individually for
each connected client.
The Context Merging architecture is also
modular but each module serves a distinct
functionality. In this architecture a Content
Aggregation module receives the context
ROBUST MULTIMEDIA TRANSMISSION OVER WIRELESS AND MOBILE NETWORKS
145
information from clients and sends it to a Context
Merging module and an Adaptation Engine module.
Based on the client information, the Context
Merging module may request new minimum size
content to the server in order to satisfy all clients’
requests, which is then adapted in the Adaptation
Engine module and delivered to the respective
clients.
The HTTP Adaptive streaming architecture is
an emerging approach, with the Adaptation Node
located in the Client, and uses a standard web
protocol for streaming of both on-demand and live
contents. The HTTP Adaptive streaming is based on
the concept of “progressive download”, but instead
of large files to download, it uses very small
“chunks” of content that can be compared to the
streaming of large packets using a conventional
streaming protocol. The contents are encoded in
many small segments (“chunks” of various sizes and
resolutions/bitrates) to a web (streaming) server that
will then receive requests from Clients. At each
Client, according to network and host conditions at
the time of real streaming (i.e., bandwidth, CPU
load, screen size, etc.), an Adaptation System
process uses several heuristics to determine the most
adequate “chunk” variant to request from the web
server.
3 SELECTING THE MOST
EFFICIENT MECHANISM
There are different options for multimedia
transmission which provide different levels of
efficiency as regards the use of resources and levels
of personalization capabilities. The use of broadcast,
though it minimizes the use of the network
resources, provides the lowest level of
personalisation capabilities, cannot be used for all
the available streams but can be applied to specific
technologies (i.e.DVB-H). The use of multicast,
though it can be deployed for a larger number of
streams than broadcasting, suffers from limitations
imposed by network providers. The advantages of
unicast when compared to the other transmission
modes is counterbalanced by its limited efficiency in
network resource utilisation, especially for large
events that interest large audiences.
Another issue to be taken into consideration is
the seamless switching between the streams. When
switching of streams happens without the need for
changing access network technology, no problem
appears, provided that the unicast and multicast
streams are be synchronized. However, if there is the
need for changing access network technology,
seamless switching cannot happen, as the decoding
modules at the terminal device deployed in the two
cases are different.
The case of switching between access
technologies for the same stream is the most difficult
one, as the transmission of the stream over the
different access technologies is not synchronized,
hence leading to potential jumping to an earlier or a
later point in the timeline or even to interruption of
the transmission for a short period until a new
connection is set up, unless IP mobility across access
networks is supported.
In case of HTTP Adaptive Streaming, switching
between access networks is much simpler, due to the
session based nature of HTTP protocol, as the
streamed information is transmitted by independent
“chunks” of the content and it is the responsibility of
Client to request those “chunks”. Therefore, the user
may experience eventual transient quality
degradation (due to different access network
characteristics), but no stream interruption.
Another important issue is the transition between
different transmission modes, namely unicast,
multicast and broadcast. Since the streams in each
mode may not be synchronised (even if they are,
buffering can lead to time-shifting) the problem of
jumping to another point in the video stream or
interruption can still happen. Use of HTTP Adaptive
Streaming does not solve the problem in this case, as
it cannot be deployed for multicast or broadcast,
making it harder to be considered a universal
solution.
Another important requirement is the existence
of mechanisms for monitoring the bandwidth
capabilities and the QoS over the different networks
by including heuristics in the Client for the
determination of the best available access network
connection.
Therefore, in order to provide a policy for the
selection of the most appropriate access technology
and corresponding access method to the transmitted
stream, the following information should be taken
into account:
Which devices support DVB-H?
Is multicast supported by the user access
network (this applies to wired and wireless IP
networks)?
What need is imperative: the need for
personalization or the need for resource usage
efficiency?
What selection offers the best quality of
experience to the user?
WINSYS 2010 - International Conference on Wireless Information Networks and Systems
146
4 CONCLUSIONS
The most suitable techniques for robust media
transmission of media streams, ensuring adequate
QoS control over wired, wireless and mobile
networks, with respect to the equipment (server or
terminal oriented) and the corresponding protocol
layer, is depicted in Figure 3.
Dynamic
QoSStream
adaptation
Selectionof
transmission
mode
Dynamic
switching
between
transmission
modes
Applicationlayer
MAC‐Datalinklayer
Serverside
Terminalside
Trafficload
balancein
IEEE802.11
networks
Network
layers
IPNetworklayer
Application+Networklayer
Figure 3: Positioning of each technique with respect to
protocol and equipment.
These techniques can be used in parallel or
individually in the final deployment of the platform
and according to the capabilities supported by the
end-to-end media distribution architecture.
Special attention should be given to the use of
HTTP Adaptive Streaming, as this technique, apart
from supporting robust media transmission with
respect to the best available QoS that can be offered,
incorporates a series of advantages. Being a
technique implemented at the application layer and
having the server providing different encoding
bitrates allows its use over a variety of terminal
devices, including mobiles. Furthermore, the use of
standards HTTP protocol introduces the advantage
of web based information exchange (firewall block
avoidance, use of standard HTTP proxies, use of
TCP congestion control mechanisms).
ACKNOWLEDGEMENTS
The authors my-eDirector 2012 project partners for
their support in the preparation of this paper.
REFERENCES
Hutter, A., Amon, P., Panis, G., Delfosse, E., Ransburg,
M., Hellwagner, H., 2005. Automatic adaptation of
streaming multimedia content in a dynamic and
distributed environment. In ICIP 2005.
Cellatoglu, A., Fabri, S., Worrall, S., Sadka, A., Kondoz,
A. 2001. Robust header compression for real-time
services in cellular networks. In 3G Mobile
Communication Technologies 2001
Patrikakis, Ch., Papaoulakis, N. 2010. Deliverable D2.4:
System Architecture and Specifications. ICT Project
My-e-Director 2012 - Real-Time Context-Aware and
Personalized Media Streaming Environments for
Large Scale Broadcasting Applications.
Giannoulis, S., Antonopoulos, C., Topalis, E.,
Athanasopoulos, A., Prayati, A., Koubias, S. 2006.
TCP vs. UDP Performance Evaluation for CBR
Traffic On Wireless Multihop Networks. In Proc of
CSNDSP 2006.
Larzon, L.-A., Degermark, M., Pink, S. 1999. UDP Lite
for Real Time Multimedia Applications. HP.
Rajamani, R., Kumar, S., Gupta, N. 2002. SCTP versus
TCP: Comparing the Performance of Transport
Protocols for Web Traffic. Computer Sciences
Department. University of Wisconsin-Madison.
Mohan, R., Smith, J., Sheng Li, C. 1999. Adapting
multimedia internet content for universal access. In
IEEE Transactions on Multimedia. 1:104–114. 1999.
Kohler, E., Handley, M., Floyd, S. 2006. Datagram
Congestion Control Protocol. RFC 4340. IETF.
Patrikakis, Ch., Papaoulakis, N., Stefanoudaki, Ch.,
Nunes, N. 2009. Streaming content wars: Download
and play strikes back. In Personalization in Media
Delivery Platforms Workshop. Venice, Italy.
Zambelli, A. 2009. IIS Smooth Streaming Technical
Overview. Microsoft.
Speakman, T., Crowcroft, J., Gemmell, J., Farinacci, D.,
Lin S., Leshchiner, D., Luby M., Montgomery T.,
Rizzo, L., Tweedly A., Bhaskar N., Edmonstone R.,
Sumanasekera R., Vicisano L. 2001. PGM Reliable
Transport Protocol Specification. RFC 3208. IETF.
Zhou, B., Fu, C.P., Lau , C.T., Foh, C.H. 2007. An
Enhancement of TFRC over Wireless Networks. In
IEEE Wireless Communications and Networking
Conference - WCNC 2007.
Stewart, R., Xie, Q., Morneault, K., Sharp, C.,
Schwarzbauer, H., Taylor, T., Rytina, I., Kalla, M.,
Zhang, L., Paxson, V. 2000. Stream Control
Transmission Protocol.RFC 2906. IETF.
Floyd, S., Kohler, E. 2006. Profile for Datagram
Congestion Control Protocol (DCCP) Congestion
Control ID 2: TCP-like Congestion Control. RFC
4341. IETF.
Jacobson, V., Braden, R. Borman, D. 1992. TCP
Extensions for High Performance. RFC 1323. IETF.
Santos, H., Cruz, R., Nunes, M. 2009. Rate Adaptation
Techniques for WebTV. In Proc. of UCMEDIA 2009.
Venice, Italy.
Cruz, R., Nunes, M., Gonçalves, J. 2009. A Personalized
HTTP Adaptive Streaming WebTV. In Proc. of
UCMEDIA 2009. Venice, Italy.
ROBUST MULTIMEDIA TRANSMISSION OVER WIRELESS AND MOBILE NETWORKS
147