UNIVERSITY OF CALIFORNIA
Los Angeles
Adaptive Multimedia in Wireless IP
Networks
A dissertation submitted by partial
satisfaction of the
requirements for the degree Doctor of
Philosophy
in Computer Science
by
Matheos Ioannis Kazantzidis
2002
@ Copyright by
Matheos Ioannis Kazantzidis
2002
The dissertation
of Matheos Ioannis Kazantzidis is approved.
______________________________
Leonard
Kleinrock
______________________________
Songwu
Lu
______________________________
Mani Srivastava
______________________________
Mario Gerla,
Committee chair
University
of California, Los Angeles
2002
“I dedicate this
thesis to my parents
Ioannis
Kazantzidis and Maria Kazantzidi/Agapaki”
TABLE
OF CONTENTS
1 Introduction. 1
2 Background
and Related Work. 6
2.1 Application
Context 6
2.1.1 Adaptability
in Multimedia Streaming. 8
2.2 Feedback. 16
2.3 Transport
and Congestion Control 18
2.4 Networks. 25
2.5 Performance
Evaluation. 26
2.5.1 Platforms. 26
2.5.2 Accuracy
of Measurement versus QoS. 27
2.5.3 Metrics. 30
3 Network
Technology. 36
3.1 Bluetooth
Simulation Model 44
3.1.1 Scatternets
and Inter-Piconet Scheduling Simulation. 45
3.2 Bluetooth
Scatternet Architecture. 47
3.2.1 Background. 48
3.2.2 Validation. 52
3.2.3 Rendezvous
Points Allocation Schemes. 52
3.2.4 Scatternet
Model & Results. 54
3.2.5 Conclusion. 63
4 The
Network Feedback Architecture. 66
4.1 Propagation
of Available bandwidth. 69
4.2 QM-AODV
QoS propagation and aggregation support 70
4.3 802.11
network layer available bandwidth support 77
4.3.1 A
Network Layer Implementation. 82
4.3.2 Simulation
Correctness. 86
4.3.3 Adaptive
Multimedia Simulation. 88
4.3.4 Call
Admission Simulation. 89
4.3.5 Conclusion. 90
4.4 The
Bluetooth Available Bandwidth Support 91
5 The
End-to-End RTP based Architecture. 93
5.1 Development
of an Adaptive Audio Client/Server 94
5.1.1 Speech
Recognition Extensions. 94
5.1.2 Testbed
and Environment 98
5.2 Real
802.11 / Wavelan experiments. 100
5.2.1 CSMA
Speech Scheme Real Testbed Results. 103
5.2.2 Issues
in Payload adaptation. 108
5.2.3 Larger
Scale Real Testbed Experiments. 109
5.2.4 Intra-Protocol
Fairness. 112
5.2.5 Tcp
and Udp Experiments with hybrid simulator 121
5.2.6 Conclusions. 123
5.3 Adaptive
Multimedia in Bluetooth Piconets. 124
5.3.1 Video
and TCP. 128
5.3.2 Voice. 129
5.3.3 Adaptive
Video and TCP. 131
5.3.4 Conclusion. 132
5.4 Bluetooth
Scatternet End-to-End Adaptation. 133
5.4.1 Results. 135
5.5 Ad-hoc
Bluetooth and 802.11 Comparison. 139
5.5.1 Conclusions. 145
6 Available
Bandwidth. 147
6.1 Packet
Pair Background. 147
6.2 Extension
of Packet Pair/Train For Available Bandwidth Sampling. 150
6.2.1 Why
do we need a new method to calculate the samples?. 150
6.2.2 The
ab-probe model 151
6.2.3 The
“bytes over time” model 153
6.2.4 Observability
and robustness of ab-probe. 155
6.3 Experiments. 158
6.3.1 Active
measurement 160
6.3.2 Passive
Measurement 165
6.3.3 Wireless
Link Measurement 168
6.4 MMTP. 170
6.4.1 Filtering. 171
6.4.2 Confidence
based Weighting Algorithm.. 172
6.4.3 Stability
and TCP Friendliness. 173
7 Comparisons. 176
7.1 802.11
single hop. 178
7.1.1 QoS. 179
7.2 802.11
multi-hop. 182
7.2.1 QoS. 184
7.2.2 Mobility. 187
7.3 Bluetooth
Scatternets. 190
7.3.1 QoS. 192
7.4 TCP
Friendliness. 193
8 Conclusion. 196
LIST OF FIGURES
Figure
2.1. The two functions of multimedia adaptation. 8
Figure 2.2 Application Context 9
Figure 2.3: General Server Architecture. 15
Figure 2.4:General Client Architecture. 16
Figure 2.5 Mean of Accuracy Normal(mean, 0.1) versus QoS. 30
Figure 2.6. Accuracy Variance versus QoS. 30
Figure 2.7 QoS MPQM related evaluation model used. 35
Figure 3.1 A scatternet with one inter-piconet unit that
divides its time between the two piconets. 44
Figure 3.2 Scatternet Simulator and Interface with NS and
GlomoSim Bluetooth model 47
Figure 3.3 The number of nodes active in the polling cycle
during the superframe cycle of a piconet with s non-gateway slaves and 3
gateway nodes. 51
Figure 3.4 The topology for the validation of the equation. 54
Figure 3.5 Average connectivity degree versus PG (x axis)
and GP (y axis) limits. 58
Figure 3.6 Round Robin f.t. without gateway priority,
RV-maxmin and RV-Separation : Minimum averaged F.T. against GP (x axis) and PG
(y axis) limit 58
Figure 3.7 Round Robin f.t. without gateway priority,
RV-maxmin, RV-greedy, RV-tree : Minimum averaged F.T. against GP (x axis) and
PG (y axis) limits. 58
Figure 3.8 . Round Robin f.t. without gateway priority,
RV-maxmin and RV-Separation : Averaged F.T. against GP (x axis) and PG (y axis)
limits. 59
Figure 3.9 . Round Robin f.t. without gateway priority,
RV-maxmin, RV-greedy and RV-tree: Average F.T. against GP (x axis) and PG (y
axis) limits. 59
Figure 3.10 Piconet pair F.t. distribution for 123
piconets (2,3) (GP,PG). 60
Figure 3.11 . Piconet pair F.t. distribution for 123
piconets (3,3) (GP,PG). 60
Figure 3.12 Piconet pair F.t. distribution for 123
piconets and no imposed limits - (7,7) (GP,PG). 61
Figure 3.13 Hop distance averaged over all piconet pairs
in scatternet. (x axis is the GP limit) 62
Figure 3.14 Hop distance distribution for the (3,3)
(GP,PG) limit (x axis is hop distance, y axis is number of piconet pairs that
are connected with the hop distance with a min-hop route) 62
Figure 3.15 End-to-end route throughput averaged over all
piconet pairs for the RV-maxmin case. (x axis is the GP limit, y axis is the
route mimimum f.t. ) 63
Figure 3.16 Route throughput distribution for the (3,3)
(GP,PG) limit (x axis is route throughput in bps, y axis is an increasing
number that represents a piconet pair, it is ordered by route throughput) . 63
Figure 4.1 Node block diagram architecture for network
feedback support. 68
Figure 4.2 The network feedback case: accurate measurement
and better propagation. 69
Figure 4.3 A scenario of AODV QoS value aggregation and
propagation. 77
Figure 4.4 (top) a CBR source uses different packet. (mid)
packet delays (bottom) the throughput observed per packet. 79
Figure 4.5 Under low load the throughput measurement
becomes independent of packet size by substracting a constant c 80
Figure 4.6 Window operation. 81
Figure 4.7 The 802.11 measurement block diagram.. 82
Figure 4.8 An example packet arrival/ acknowledgement
sequence. The network calculates the contributed delays by using the
notification series to emulute the link queue. 85
Figure 4.9 Exact measurements for CBR.. 87
Figure 4.10 The VBR source rates in time averaged over 32
packets. 87
Figure 4.11 Accurate measurements for VBR traffic. 87
Figure 4.12 Overall loss rates (%) per experiment 89
Figure 4.13 Total bytes sent vs total bytes received with
call acceptance and without. 89
Figure 5.1 The topology used in the experiments. Three
subnets are used to create a multihop connection. Two subnetworks are used for
the server and the client respectively. One is used for the interference. The
gateway has virtual interfaces to all subnets. 98
Figure 5.2 : Loss Rate when no adaptation is performed,
and with standard experiment interference. Audio stream is packetized in 960
byte packets and sampled at 22000 samples per second and 8 bit per sample.
Interference is ‘ping’ packets of 40 bytes attempting to fill the channel 103
Figure 5.3 Visualization of text synthesizer use:
switching between audio (even number) and text stream (odd number) with
different TTS-thresholds in first experiment 104
Figure 5.4 : Loss rates when adapting to QoS, and with
standard experiment interference. Audio stream is packetized in 240 byte
packets and sampled at 8000 samples per second and 8 bit per sample (due to the
adaptation) . Interference is ‘ping’ packets of 40 bytes. 106
Figure 5.5 Visualization of text synthesizer use:
switching between audio (even number) and text stream (odd number) with
different TTS-thresholds insecond experiment 106
Figure 5.6 : Loss rates when adapting to QoS with
extremely adverse network conditions. Audio stream is packetized in 240 byte
packets and sampled at 8000 samples per second and 8 bit per sample.
Interference is ‘ping’ packets of 160 bytes attempting to fill the channel. The
distance between the stations here is larger and the larger propagation delays
result in higher loss rates. 107
Figure 5.7 Visualization of text synthesizer use:
switching between audio (even number) and text stream (odd number) with
different TTS-thresholds in third experiment 107
Figure 5.8 Topology in Large Scale Real Experiments. 110
Figure 5.9 Loss Rates in Clients with FTP interfering
traffic. 110
Figure 5.10 Loss Rates in Non-Adaptive Clients with FTP
interfering traffic. 110
Figure 5.11 Loss Rates in Clients with only non-TCP
intefering traffic. 111
Figure 5.12 A single hop link ping delay graph. 112
Figure 5.13 Adaptation mechanism.. 115
Figure 5.14Averaged client loss rates vs adaptivity. 117
Figure 5.15 Averaged effective bandwidth vs adaptivity. 118
Figure 5.16 Averaged server consumed bandwidth vs
adaptivity. 118
Figure 5.17 Averaged Coefficient of Variations in
Effective Client Bandwidth. 119
Figure 5.18 Coefficient of Variation of Effective
Bandwidth among clients (y axis) for 100 generated topologies (x axis) when
TCP/FTP traffic exists. 121
Figure 5.19 Coefficient of variation of client rates of
the 10 connections (y axis) for all 100 generated topologies (x axis) with
802.11 MAC and no mobility. 122
Figure 5.20 Coefficient of variation of client rates of
the 10 connections (y axis) for all 100 generated topologies (x axis) with
802.11 MAC and mobility. 122
Figure 5.21 Coefficient of variation of client rates of
the 10 connections (y axis) for all 100 generated topologies (x axis) with CSMA
MAC and no mobility. 123
Figure 5.22 Coefficient of variation of client rates of
the 10 connections (y axis) for all 100 generated topologies (x axis) with CSMA
MAC and mobility. 123
Figure 5.23 A few seconds from the H263 source trace (sec,
bytes) 126
Figure 5.24 Bluetooth end to end adaptation. 128
Figure 5.25 H.263 Non adaptive video and TCP connections
aggregate throughput. 129
Figure 5.26 Loss Rates for video connections for H.263. 129
Figure 5.27 Voice Delay Distribution for WaveLAN.. 130
Figure 5.28 Voice Delay Distribution for Bluetooth. 130
Figure 5.29 H.263 aggregate server sent rates. 130
Figure 5.30 Loss Rates for Adaptive H.263 Video. 131
Figure 5.31 H.263 adaptive video and TCP connections
aggregate throughput. 132
Figure 5.32 The canonical recursive Bluetooth scatternet
topology. 135
Figure 5.33 Per Connection Server Send Throughput and
Goodput for Bluetooth and 802.11, adaptive and non-adaptive for the 4 by 4
piconet case. 137
Figure 5.34 Loss rates. 137
Figure 5.35 Jitter and delays for a typical connection.
Left: from the 3 by 3 piconets case, Right: From the single piconet case. 137
Figure 5.36 Adaptive connections in the (top) 4 by 4,
(middle) 3 by 3 and (bottom) 2 by 2case. Left column shows (top) reported RTP
loss rates (middle) layer from which frame is picked for transmission (bottom)
received rate averaged over 1 second. Right column shows jitter and delay. 142
Figure 5.37 The corresponding 802.11/DCF exact topology
cases (top, left) 64 node, eq. to 4 by 4 (top,right) 36 node, eq. to 3 by 3
piconets (bottom, left) 16 node, eq. to 2 by 2 and (bottom, right) 4 nodes,
eq. to 1 piconet Bluetooth case. 143
Figure 5.38 Direct Comparison of Bluetooth and 802.11 MAC
using an end-to-end adaptation mechanism. (top) Average Recv throughput, Sent
Throughput (middle) Loss rates, Delay Jitter (bottom) Delay, Packets Delivered 144
Figure 6.1 Possible events in links before the bottleneck
link. 147
Figure 6.2 Possible events at the bottleneck link. 148
Figure 6.3 Events occurring after bottleneck link. 148
Figure 6.4 ab-probe model 149
Figure 6.5 “bytes over time” available bandwidth
calculation varies with train size while network conditions remain constant. 154
Figure 6.6 The “bytes over time” relative error for a
4Mbps link with 4Mbps up to 1.280Mpbs available 155
Figure 6.7 The “bytes over time” relative error for a
4Mbps link with 1.230Mbps to 100Kbps available 155
Figure 6.8 How an error in Pb estimation affects the
available bandwidth sampled with ab-probe. 157
Figure 6.9 Difference between “bytes over time” and
ab-probe measurement. 157
Figure 6.10 The short range campus Internet experiment
topology. 159
Figure 6.11 The long range campus Internet experiment
topology. 159
Figure 6.12 The sender bandwidth of the MPEG-4 video. 160
Figure 6.13 Active Measurement using BOT and AB-probe in
the short range Internet topology. 163
Figure 6.14 Active Measurement using BOT and AB-probe in
the long range Internet topology. 163
Figure 6.15 . Graphs from the short distance active
measurement experiment a. Ab-probe short distance packet pair samples when
injected traffic is 0Mbps b. BOT packet pair samples when injected cross
traffic is 0Mbps c. Ab-probe short distance packet pair samples w when injected
cross traffic is 3Mbps e. Ab-probe packet pair samples filtered f. BOT packet
pair samples filtered. The measurement is not affected by the injected traffic
in all cases. 164
Figure 6.16 Graphs from the long distance active
measurement experiment. a. Ab-probe short distance packet pair samples when
injected traffic is 0Mbps b. BOT packet pair samples when injected cross
traffic is 0Mbps c. Ab-probe short distance packet pair samples when injected
cross traffic is 3Mbps e. Ab-probe packet pair samples filtered f. BOT packet
pair samples filtered. The measurement is not affected by the injected traffic
in all cases. 165
Figure 6.17 Model diversion graph for pairs and trains for
the short distance case. 167
Figure 6.18 Model diversion graph for pairs and trains for
the long distance case. 167
Figure 6.19 Graphs from the short distance active
measurement experiment. a. Ab-probe short distance packet pair samples when
injected traffic is 0Mbps b. BOT packet pair samples when injected cross
traffic is 0Mbps c. Ab-probe short distance packet pair samples when injected
cross traffic is 3Mbps d. BOT packet pair samples when injected cross traffic
is 3Mbps (note that the samples may look more than the ones in b. but in
reality they are just more scattered) 168
Figure 6.20 Graphs from the wireless measurement
experiment. a. Sender timestamps of FTP traffic when injected traffic is 0Mbps
b. Sender timestamps of FTP traffic when injected traffic is 3Mbps c. Ab-probe
packet pair samples when injected cross traffic is 0Mbps d. BOT packet pair
samples when injected cross traffic is 0Mbps e. Ab-probe packet pair samples
when injected cross traffic is 3Mbps f. BOT packet pair samples when injected
cross traffic is 3Mbps 168
Figure 6.21 Wireless link measurement 169
Figure 7.1 Topology used in the comparison experiments. M
denotes a node that would be a Master in the Bluetooth scenario, S a slave. A
direct line between a Slave and a Master indicates a Slave’s Home Piconet, a
dotted one that the slave acts as a gateway to another piconet 177
Figure 7.2 Loss Rates. 179
Figure 7.3. QoS, 802.11 single hop. 180
Figure 7.4 Loss Rates, 802.11 single hop, higher rates. 181
Figure 7.5. QoS, 802.11 single hop, loss rates. 182
Figure 7.6 Loss rates on 802.11 multihop. 184
Figure 7.7 QoS on 802.11 multi-hop. 185
Figure 7.8 Loss rates in 802.11 multihop with higher rate
video. 186
Figure 7.9 QoS in 802.11 multihop with higher rate video. 187
Figure 7.10 5km/h mobility. 188
Figure 7.11 20km/h mobility. 188
Figure 7.12 55km/h mobility. 189
Figure 7.13. QoS in 5 km/h mobility. 189
Figure 7.14 QoS in 20 km/h mobility. 190
Figure 7.15 QoS in 55km/h mobility. 190
Figure 7.16 Bluetooth Scatternets Loss Rates. 192
Figure 7.17 Bluetooth QoS. 193
Figure 7.18 TCP Friendliness Experiment 195
Figure 8.1 Middleware and Agent techniques expected
operation. 199
LIST OF TABLES
Table 4‑1 Pseudo code for QM-AODV modifications to
AODV.. 75
Table 4‑2 Sample Run of QM-AODV operation. 76
Table 5‑1 Real Testbed parameters. 100
Table 5‑2 Adaptation mechanism parameter values. 115
Table 5‑3 Configurations tested. 115
Table 5‑4 Simulation Parameters. 117
VITA
November 12, 1971 Born, Athens, Greece
1995 Diploma
of Higher Education
University of Patras,
Patras, Greece
1995 Computer
Engineer
Advanced
Informatics Ltd
Patras, Greece
1996-1998 Teaching
Assistant
University
of California
Department of Computer Science
Los Angeles, California
1998 Recipient
of The Gerondelis Foundation Fellowship
1998 M.S.
Computer Science
University
of California
Los Angeles, California
1998-2002 Research
Assistant
University
of California
Department of Computer Science
Los
Angeles, California
2002
Recipient of the Fred W. Ellersick Prize
Formerly known as
The Communications Society Magazine Prize Paper Award
PUBLICATIONS AND PRESENTATIONS
Kazantzidis M., Chen T., Romanenko Y., Gerla M.
(March, 1999) An Ultimate Encoding Layer for Layered Real-Time Speech Streams
over Multi-hop Wireless Networks. Proceedings of IEEE 2nd Annual Conference on
Wireless Comm., San Diego, CA
Kazantzidis M., Tang K., Gerla M. (May, 1999)
Validation of Multi-Layer Simulation Experiments via Analysis and Measurements.
Proceedings of DARPA/NIST Network Simulation Workshop, Fairfax, VA
Kazantzidis M., Slain I.,
Chen T., Romanenko Y., Gerla M. (June, 1999) Experiments on QoS Adaptation for
Improving End user Speech Perception over Multihop Wireless Networks.
Proceedings of IEEE ICC, Vancouver, Canada.
Kazantzidis M., Wang L., Gerla M. (November,
1999) On Fairness and Efficiency of Adaptive Audio Application Layers for
Multihop Wireless Networks. Proceedings of IEEE MOMUC'99, San Diego, CA
Kazantzidis M. (1999) Increasing Speech
Perception in face of external interference - Secretary of the Army Louis
Caldera UCLA visit
Gerla M., Kazantzidis M., Pei G., Talucci F., and
Tang K. (2000) Ad Hoc, Wireless, Mobile Networks: The Role of Performace
Modeling and Evaluation. Book Chapter In Performance Evaluation: Origins and
Directions, pp. 51-95, Edited by G. Haring, C. Lindemann, and M. Reiser,
Springer-Verlag, 2000.
Kazantzidis M. (September 2000) Adaptive Video
over Multi-Hop Wireless Networks using Hybrid Simulation – Demonstration
Presentation at Digivations 2000, Digital Media Innovation Program, Santa
Barbara CA
Kazantzidis M. (February 2001) Wireless Adaptive
Multimedia using Network Measurements. UCLA Computer Science Technical Report #200102
Kazantzidis M., Lee S.J., Gerla M. (2001)
Permissible Throughput Network Feedback for Adaptive Multimedia in AODV
MANETs – Proceedings of IEEE ICC 2001
Kazantzidis M. (2001) How to measure available
bandwidth on the Internet. UCLA Computer Science Technical Report #010032
Kazantzidis M. (2001) Locally optimal Bluetooth
Scatternet formation. UCLA Computer Science Technical Report #010033
Kazantzidis M. (2001) End-to-end versus Explicit
Feedback Measurement in 802.11 Networks. UCLA Computer Science Technical Report
#010034
Gerla M., Kapoor R., Kazantzidis M., Johansson P.
(2001) Ad hoc Networking with Bluetooth. Wireless Mobile Internet
Conference during MobiCom 2001
Johansson P., Kapoor R., Kazantzidis M., Gerla M.
(October 2001) Bluetooth an Enabler
of Personal Area Networking. IEEE Network Special Issue on Personal
Area Networks, Sept-Oct 2001
Kazantzidis M., Zanella A., Gerla M. (2002)
End-to-end Adaptive Multimedia over Bluetooth Scatternets Proceesings of Eurel
AICA European Wireless Conference 2002
Johansson P., Kapoor R., Kazantzidis M., Gerla M.
(2002) Rendezvous Scheduling in Bluetooth Scatternets – Proceedings of IEEE ICC
2002
Johansson P., Kapoor R., Kazantzidis M., Gerla M.
(2002) Personal Area Networks:
Bluetooth or IEEE 802.11? International Journal of Wireless Information
Networks SPECIAL ISSUE ON MOBILE AD HOC NETWORKS (MANETs) Standards, Research,
Applications, April 2002
Kazantzidis
M., Gerla M. (2002) End-to-end versus Explicit Feedback Measurement in 802.11
Networks. Proceedings of the 7th IEEE
Symposium on Computers and Communications
Kazantzidis
M., Gerla M. (2002) On the Impact of Inter-Piconet Scheduling in Bluetooth
Scatternets – Proceedings of WWIC 2002
ABSTRACT OF THE DISSERTATION
Adaptive Multimedia in Wireless IP
Networks
by
Matheos Ioannis Kazantzidis
Doctor of Philosophy in Computer
Science
University of California, Los
Angeles, 2002
Professor Mario Gerla, Chair
Support for video and audio applications is important to
single and multi hop wireless networks whether they are used as extensions to
the Internet or not. Due to the variability of response of the air medium and
the mobility support that is expected of these networks, it is accepted that
such applications must dynamically adapt to network conditions, taking
advantage of the different content representations achieved by advances in
coding. This adaptability targets at maximizing the overall QoS delivered by
the network and may be classified into (i) The transport functionality that
decides the network parameters e.g. sending rate and (ii) The presentation
functionality that decides the content that should fit the network parameters.
In wireless, it is particularly difficult to implement an accurate monitoring
process (measurement) and embed it into a distributed strategy that efficiently
controls the scarce network resources. Therefore, transport protocols designed
for wired networks fail. This, combined with scalability challenges of some ad
hoc environments (e.g. battlefield) motivates the exploration of an important
trade-off for this environment. On the one hand, adopting a thin and scalable
network architecture allows for end-to-end adaptation which is limited in
measurement accuracy and consequently performance. On the other hand, the implementation
of lower layer feedback support leads to architectures that are less scalable
and bear a higher deployment cost. But, can deal effectively with the
measurement inaccuracy problem. In order to explore this tradeoff, we study,
develop and improve existing end-to-end as well as network feedback strategies
in terms of the overall QoS delivered to the network users. We trade-off the
end-to-end techniques versus the network feedback techniques and provide a
performance gain model that can guide the design of real time applications as
well as the design of the networks to support them.
A
great deal of work is targeted at exploiting adaptive mechanisms in all design
layers of wireless networks. The goal is to gain the desired protocol responsiveness
that deals with the frequent unexpected changes in grades of service. The air
medium and the mobility support expected of both last hop wireless internet and
ad-hoc multi-hop wireless networks requires careful and specialized higher
layer protocols for congestion control and QoS support. The ones developed for
wired networks fail when put to work in a wireless environment.
The
responsibility for flow control on the Internet is mainly left at the transport
layer, allowing for a scalable design and a thin network layer. The transport
peers perform some type of monitoring to its packets and apply sampling and
estimation techniques to calculate desirable quantities e.g. trip times, path
available bandwidth etc. Explicit help from lower layers is not allowed, as
this would impair scalability, make deployment difficult and dramatically
increase cost. TCP for example, is using a flow’s single packet loss as an
indication of network congestion, presuming that the packet is dropped due some
stressed buffer along its path. This however does not work in wireless
networks, as packet losses due to external and internal interference are also
frequent.
We do not directly deal with TCP in
this thesis, as it has been shown to be an unsuitable protocol for multimedia
communication, but we frequently refer to it and consider part of the work
applicable for TCP protocols. Even if reliability (ARQ and re-ordering)
functionality is removed from TCP, it still does not present a good candidate.
The trial-and-error approach converges slowly and requires many attempts. On
the other hand, multimedia traffic prefers an approach that would, from
scratch, operate on a fairly good estimate of the available bandwidth, incurs
minimum perceptually costly attempts to improve quality, and adjusts smoother
to network changes. Furthermore, window based techniques impose unnecessary
delays and high jitter. Real time traffic should ideally be serviced
(transmitted) as soon as generated by the application, or packets may not reach
the destination by their playback time. Live applications are particularly
intolerant of delays, especially transport delays that are always in the
critical path.
Adaptive
multimedia transports would ideally prefer to have accurate knowledge of the
bandwidth available along their path, averaged over a small interval. Let us
define the available bandwidth over one link as the link bandwidth minus the
used bandwidth, i.e. the un-utilized bandwidth. The path’s available bandwidth
then would be the minimum available bandwidth across all links in the
end-to-end path. With this information at hand, the peers would be able to
adjust their rate so as to minimize their lost, not played packets, perform
congestion control and be TCP fair or friendly. At the same time it is hard to
measure available bandwidth, especially in an end-to-end fashion. It is a
highly variable quantity and constrained to an end-to-end observation, as the
Internet Protocol scalable architecture dictates. Current available bandwidth
techniques have been developed for wired networks, and have to approximate the
network as performing weighted fair queuing on its flows [Pax97]. The Internet,
however, cannot distinguish flows, may employ a variety of queuing disciplines
and currently has pre-dominantly FIFO routers and therefore current
measurements are highly unreliable.
In
a nutshell, current end-to-end transport solutions for multimedia communication
are largely heuristic and allow significant room for improvement. The same, put
to work in wireless networks, are unexplored and non-promising. Their variant
response and heterogeneity places even more stringent requirements in sampling,
filtering methods and convergence times. Therefore, besides special development
of end-to-end methods another option becomes particularly worth exploring, i.e.
deploying network support for transports and applications. Let us call such
architectures, network feedback architectures.
Such
architectures require special node support, possibly in both hardware and
software. Each node measures its bandwidth and delay performance. This can be
done fairly accurately because lower layers perform it, each knowing their own
mechanisms. The values are then propagated using routing or other similar
protocols. Eventually they reach the end-hosts where they may be used by
transports, applications or measurement based call admission algorithms. Such a
setting has the advantage that it may overcome the aforementioned difficulties
improving the overall performance and QoS. However, since each node requires
special support, deploy-ability, scalability, inter-operability and
consequently cost are impaired. Note that per flow QoS is not required, just a
per link QoS information estimation and a per routing table entry aggregate
variable per QoS metric. Given the bad performance of existing end-to-end
techniques and that those requirements vary in wireless networks, it is
particularly worth exploring such architectures.
In particular this work deals with
the wireless technology to be used for the deployment of personal area and
multi-hop networks in ad-hoc as well as Internet extension settings. Namely
PANs and MANETs of Bluetooth and 802.11. In Bluetooth both Piconet and
Scatternet configurations are of interest. In 802.11 we look at multi-hop and
single hop configurations. These two types of wireless networks have a very
different philosophy in their medium access control. Bluetooth organizes the
nodes into centrally controlled groups called Piconets while 802.11/DCF assume
a totally distributed access control so that mobility support is more flexible.
These two approaches provide the two basic MAC options over which we explore
the measurement accuracy and application adaptation.
We find that existing methods of
end-to-end adaptation reduce the application loss rates according to the
offered load. This means that it is possible to have QoS improvement when using
these techniques over multi-hop wireless networks. However, using present state
CODEC technology and rates the above condition is finally true only for small
networks and low number of adaptive connections.
We improve end-to-end techniques by
using an improved packet dispersion technique that is based on an innovative
and intuitive sampling method, other than the ‘bytes-over-time’ which has been
extensively used in the past. We show that this method works better than other
end-to-end methods. While it pushes the end-to-end limits on network size and
number of flows, it is still limited by the round-trip delayed feedbacks,
multi-hop measurement noise and reverse path problems.
We therefore develop the network
feedback solutions for these networks. Comprised of a per source-destination
single hop highly accurate node measurement and a QoS value propagation
technique that gets the necessary information available to the sources, it
deals very effectively with the congestion control problem of wireless
networks. It consistently exhibits increased QoS, using perceptual QoS metrics,
when compared to non-adaptive transmission.
In summary, we consider the significant
contributions of this work to be:
-
The study of the
limitations of end-to-end techniques using a perceptual QoS evaluation model
and hybrid simulation
-
The development
of an asynchronous hybrid simulation platform for Video over large multi-hop networks.
-
The development
of network feedback architectures, measurement and measurement propagation
techniques, that effectively deal with the congestion problem maximizing
perceptual QoS
-
The contribution
of practically performing locally optimal Inter-Piconet Scheduling in Bluetooth
Scatternets.
-
The simple, low
cost, Bluetooth available bandwidth measurement
-
The 802.11
available bandwidth measurement using the link ACK/LF messages
-
The Q-AODV
extension to support propagation of QoS values in parallel to the QoS routing
function.
-
The AB-probe, and
end-to-end available bandwidth measurement that is based on an innovative
sampling of packet dispersion and can be used in non-WFQ networks
Adaptation starts at the
application’s flexibility to carry on its useful task, for example meaningful
communication in a multimedia conference, in different grades of service. The
network is assumed able to support the application’s highest demand at light
load. One or more encoders may define the different grades of service by using
different compression rates. If the application or middleware is able to switch
between layers at any time, extra information needs to be maintained, either
encapsulated in an RTP-type [Rtp96] packet or in a separate packet (even
stream) indexed or synchronized to the data packet (or stream). This is because
the application needs to be aware of the received stream characteristics. This
required extra information introduces overhead. In an end-to-end architecture
the applications ability to switch between different rates has to be combined
with monitoring and quantifying the underlying network conditions. In a network
feedback architecture these are provided by the network. Since the codec
belongs to the application layer a significant part of the adaptation has to
belong there, according to the RTP paradigm.
Multimedia applications are sensitive
to lost packets, delayed packets and jitter. RTP defines how loss and jitter
should be estimated. Their monitoring is performed along the end-to-end path
and a feedback packet informs the server periodically. The server uses this
past interval to adjust its future sending rate. In essence, the underlying
assumption is that the near future network response is anticipated to be
similar to that experienced in the near past. QoS information is therefore very
time-sensitive. Another realization related to the feedback path delay is that,
sadly, when we need the QoS information the most –that is when the network conditions
are highly adverse- it is exactly when they are usually received late, errored
or lost altogether (proportional to how symmetric are the links).
By application context we refer to an
ILP architecture that implements at least an RTP thin transport layer, a presentation layer
and an application layer suitable for energy efficient, mobile clients. The
target applications for this work can be categorized as follows:
·
According to
liveliness
Live vs Playback - A source may
transmit a stream as it is captured in a live session. Or otherwise a
pre-recorded, pre-encoded stored stream is played back from secondary storage
in a playback session. The main distinction between live and playback session
is that in a live session the future data is not available whereas in a
playback all the future data is available at the beginning of the session. Thus
the stream can be transmitted arbitrarily faster (or slower) than its
consumption rate. Data is still slightly delayed and buffered at the source
before transmission to allow for a small transmission rate differences. In both
cases data has to be buffered at the client side in order to compensate for the
jitter in delivery times. Clearly a live session places more stringent
requirements in the buffer size used because use of larger buffers limits
interactivity due to the proportionally larger playback delays. Smoothing is
therefore limited too in live applications. Adaptive encoding approaches can be
used to smooth the bandwidth requirements of the encoding stream, for example
[NgK97]
·
According to
content:
Audio coding / Video coding - Audio
codecs usually require less bandwidth than video. Video codecs employ compression in two
levels. On an image level and on a frame level. Image compression is employed
to encode the image of one frame. In order to compress further usually a video
codec will transmit one compressed full significant frame and a few of the next
frames will be encoded only as differences from the significant frame. This
implies that, an audio application can usually transmit at a constant average
bit rate in a much smaller time scale than a video source. It also implies that
in video codecs packets do not have a uniform QoS significance. Furthermore,
most audio codecs have been designed for synchronous channels, for example the
GSM codec at 13Kbps. In/tolerance to lost packets is an important attribute to
audio codecs. Object coding attempts to encode audio and video (M-peg 4) by
objects. A speech application is a constant bit rate with silence, on/off intervals.
The Brady model is widely used for voice traffic modeling. A hypermedia
application is one that combines audio, video with text and images. In this
dissertation we extensively deal with Mpeg audio and video (H.263).
By
introducing adaptivity to the application layer, demands placed on the network
can be adjusted within the session, and user perception may be enhanced.
Application models that support adaptivity have been proposed in [McI98],
[Cha97], [Sis97]. An
application model suitable for wireless ad-hoc networks can be found in
[Kaz99]. Architectural considerations are found in [Cla90], [Ott98], [Boc96].
Figure 2.1.
The two functions of multimedia adaptation
We conceptually classify
the functions of multimedia adaptation in two categories: (i) the network
transport operations i.e. the network monitor that implements the measurements
and congestion control mechanisms and (ii) the presentation operations that constrained
by the network resources as reported from (i) direct the coding operations. The
challenge in (i) is to perform efficient, stable, distributed and accurate
congestion control while the challenge in (ii) is to trade-off quality and
error control for the particular content and codec capabilities.
Live and playback
multimedia applications are likely to be developed over any one of the network
configuration under study, independently of whether these will provide quality
of service or not. An example is the Internet, where shortly after the modems
reached the minimum possible bandwidth a significant number of solutions were
instantly provided and deployed. Adaptive streaming over IP networks has been
the focus of research for this reason.
In this section we
introduce an application architecture framework that will allow us to more
effectively study the application context functions.
·
Session
Initiation: This has special importance for adaptive multimedia application
protocols, since it may include QoS negotiation and call admission as well as
initial application buffering. We include both initial buffering and
re-buffering in the session initiation since it is common that many protocol
parameters are decided during this phase.
Figure 2.2 Application Context