Scalable Multiple Description Coding and Distributed Video Streaming Over 3G Mobile Networks
Scalable Multiple Description Coding and Distributed Video Streaming Over 3G Mobile Networks
Scalable Multiple Description Coding and Distributed Video Streaming Over 3G Mobile Networks
'
0 ) 1 (
0 ) 1 (
Q if AvQ W
Q if Q W AvQ W
AvQ
m
q
q q
If TH
min
AvQ < TH
max
Drop the arriving packet with probability P
Else if TH
max
AvQ
Drop the arriving packet
where
( )
ef
q
W 2 1
,
( ) ) ( 1
min max min
TH TH TH AvQ p P
,
m = queue_idle_time / transmission_time.
65
(W)RED takes advantage of the TCP retransmission mechanism. However, as
discussed in Section 3.4, UDP is more suitable for video streaming. All the random
dropping packets will be considered as packet loss and will not be retransmitted. Because
of error propagation of streaming video, the effect of packet loss gets worse. Since MPEG-
4 video is predictive inter- frame coded and layered coded, artifacts due to random packet
dropping can persist for many frames or layers. For example, consider a 30 frame/s MPEG
video sequence with one I frame every 15 frames. If an error occurs while transmitting the I
frame, the effect persists for 15 frames, or 500 ms, which is quite noticeable to a viewer.
Jill and Gaglianello analyzed and presented the results of the relationship between the
packet loss rate and the frame error rate [35], shown in Figure 4-7, from a study of
streaming MPEG compressed video over the public Internet, using the RTP and UDP
transport protocols. Similarly, if an error occurs while transmitting the base layer, its
enhancement layers have to be discarded. It means that stochastically isolated single packet
loss or bit error is converted to burst packet loss or bit errors. Therefore, early random
packet dropping before congestion is not suitable for video or audio streaming.
Figure 4-7 Packet loss effect on frame error rate
A novel marking algorithm for MPEG-4 video encoded by SMDC is proposed in
Table 4-3 to support the priority assignment after the classification given in Table 4-2. In
the proposed SMDC scheme, we re-organize different types of information, such as
shape, motion, and texture, in the bit-stream of an object-based video into four different
layers. Moreover, different layers of information are packetized into three different
66
classes of Application Level Packets (ALP) with various priorities, so that more
important compressed information can be put into higher priority packets and less
important information into lower priority ones. In comparison with the DVMA solution
where there is only one queue with three different levels of precedence for video stream,
each AF class in the proposed algorithm has one separated queue. This algorithm can be
implemented by class-based Weighted Fair Queuing (WFQ). Details of the WFQ
algorithm can be found in [76]. WRED should be disabled in each class.
Table 4-3 Proposed IP DiffServ MPEG-4 video marking algorithm
Control
Information
Shape Base
Layer
Texture Base
Layer
Texture PFGS
Layer
Texture
PFGST Layer
OD &
BIFS
DSCP = EF
or AF11
- - - -
I-VOP -
DSCP = AF11
(Class I
stream)
DSCP = AF11
(Class I
stream)
DSCP = AF21
(Class II
stream)
-
P-VOP -
DSCP= AF11
(Class I
stream)
DSCP= AF21
(Class II
stream)
DSCP = AF31
(Class III
stream)
-
PFGST
VOP
- - - -
DSCP = AF31
(Class III
stream)
In addition, MPEG-4 introduces extra data control stream, such as the object
descriptor (OD) and scene description (BIFS). These signalling streams are very loss-
and jitter-sensitive and need to be protected and marked as EF or AF11 if EF PHB is not
available.
4.2.3 Evolution of System Model for Native IP DiffServ
Evolution of Mobile Network Model
In order to support DiffServ in UMTS, we propose to copy the DSCP value in the
inner IP header to the outer IP header at encapsulation and copy the outer header's DSCP
67
value to the inner IP header at decapsulation. This mechanism allows GTP/FP tunnels to
be configured without regard to DiffServ domain boundaries.
Figure 4-8 UMTS network model in IP native mode with 3G AR
A more efficient alternative, however, is to keep the native DiffServ processing
procedure in the mobile network rather than the modified one as above (i.e., copying
DSCP value). This requirement leverages the evolution of UMTS towards its final all IP-
based phase, which is depicted in Figure 4-8. Note that the IP-based CN has enlarged to
the edge of UTRANs compared with the network model shown in Figure 2-2. The
functions of the GGSN, SGSN and RNC are further combined and implemented at one
node, Access Router (AR). The GGSN and SGSN functions within the 3G AR provide all
the UMTS-specific accounting and security features. The rest of the CN consists of
regular routers and switches that forward packets on the basis of the user-level IP
addresses. The Border Router (BR) denotes the functionality to avoid unwanted traffic
between GPRS CN and the Internet. One or more BRs are served as gateways to the
public Internet.
This network architecture provides a solution to implement the IP native mode
forwarding in a larger portion of the operators domain independent of any given access
technology, and hence can be used by the operator to support heterogeneous access
68
networks. As the coverage of the IP native mode increases, the wireless-specific
protocols are pushed farther toward the access segment. The operator may share the
domain with other access techniques by just using a specialized AR. For example, an
IEEE 802.11 AR may coexist with a 3G AR, using the same CN.
Figure 4-9 D-MDMN network model in IP native mode with 3G AR
Along with the enlargement of the IP-based CN, the introduction of D-MDMN in
the 3GPP network is depicted in Figure 4-9. The IP-based Media Delivery Network can
also be enlarged around ARs towards the whole IP-based CN, so that the media streaming
services can be pushed further to the edge of UTRANs.
Conceivably, the IP native mode coverage can be extended into the UTRAN by
implementing an AR with collocated Node B, RNC, SGSN and GGSN functions. Some
equipment vendors are adopting this approach by building what are known as intelligent
base stations with varying combined functionalities.
Evolution of Protocol Stack
In order to further analyse the implementation of DifferServ in Figure 4-9, the
protocol stack in the IP native mode for non-video services is shown in Figure 4-10 (a),
while that for video services is shown in Figure 4-10 (b).
69
Figure 4-10 Evolution of Protocol stack (Data plane)
As the coverage of the IP native mode increases, the stack becomes more efficient,
and the whole CN uses regular IP forwarding based on the end-users IP address instead
of the tunnel ID. FP frames are transported to and from the ARs over an IP network in the
transport mode. If the ARs are the next hop of Node Bs, the tunnels are short enough that
the ingresses (i.e., ARs) can execute DiffServ PHB based-on the inner IP header before
the it is encapsulated by the addition of the outer FP tunnel header, and the egresses can
also perform DiffServ PHB based-on the inner IP header after it is decapsulated by the
removal of the outer FP tunnel header. Thus, the native DiffServ processing procedure in
the mobile network is implemented rather than copying DSCP value from the inner IP
header to the outer IP header.
70
4.3 Proposed Handoff Procedures for Video Streaming
4.3.1 Proposed intra-RAN Handoff Procedure
Under the proposed intra-RAN network model of the distributed multimedia
delivery mobile network, shown in Figure 3-3, and the same assumptions given in
Section 2.3.2, the intra-RAN handoff procedure for media streaming are proposed as
follows, which also consists of three phases. The control plane of handoff procedure is
shown in Figure 4-11, while the data plane of these three phases is shown in Figure 3-3.
Only the differences are described in comparison with that in UMTS Release 4.
Figure 4-11 Propose intra-RAN handoff procedure (Control plane)
71
l Phase I: Preparation of RNS handoff and resource allocation
The control plane of the proposed handoff phase I is almost the same as that in
UMTS Release 4, shown in Figure 2-4, except that the current position (offset) of the
received media stream should go along with measurement report given by the MS. It is
the only state information required for session migration which is small enough to be
hidden inside the handoff signalling and be relayed to the SGSN.
On the data plane of handoff phase I, at the end of the preparation phase, the sRNS
stops transmitting downlink data to the MS but will not store all downlink data which
continue to arrive from the SGSN to the sRNC as no data forwarding is required.
l Phase II: Moving the Serving RNS role to target RNS
The most important difference between the proposed handoff phase II in the control
plane and that in UMTS Release 4 in Figure 2-4, is that the Stream Re-establishing takes
the place of the Media Stream Forwarding. There are no buffered data required to be
forwarded. As soon as the GTP tunnel is created between the tRNS and the SGSN, the
SGSN initiates the MD-request message (signal # 7) and the uplink flow is switched from
the old path to the new path. Upon receiving the MD-request, the set of MDSs
surrounding the SGSN starts the downlink media delivery from the offset point of the
same stream at the handoff decision according to subscriber-service bindings in VLR
(i.e., how many descriptions and layers the MS subscribes). In other words, the media
stream is re-established. The MD-request message contains the offset information at the
handoff decision point.
l Phase III: Releasing resource reservation in the old path
The most important difference between the proposed handoff phase III and that in
UMTS Release 4, shown in Figure 2-4, is that there is no buffer requirement in the BSs
for data forwarding and resequencing. Only a smaller buffer is needed in the BSs for
absorbing the delay jitter of a video stream and for re-ording due to changes in routing
paths. The functionality of multiple description assembly is implemented in the MSs.
4.3.2 Proposed inter-RAN Handoff Procedure
Under the proposed inter-RAN network model of the distributed multimedia
delivery mobile network, shown in Figure 3-4, and the same assumptions given in
72
Section 2.3.2, the proposed inter-RAN handoff procedure for media streaming also
consists of three phases. Here the control plane of handoff procedure is briefly presented
in Figure 4-12, while the data plane of these three phases is shown in Figure 3-4.
l Phase I: Preparation of RNS handoff and resource allocation
Note that there is no GTP tunnel required between SGSNs compared to UMTS
Release 4, since no data forwarding is required.
l Phase II: Moving the Serving RNS role to target RNS
l Phase III: Releasing resource reservation in the old path
Figure 4-12 Proposed inter-RAN handoff procedure (Control plane)
73
4.3.3 Handoff Enhancement for Streaming Services
The advantages of the proposed handoff approach for media streaming are
summarized as follows:
1. Due to the replacement of stream re-establishing with data forwarding, the
handoff latency can be reduced, which also reduce the buffer size in the BSs.
In the UMTS Rel 4 handoff procedures, for downlink media streams, there are two
possible situations when media stream gap or overlapping may happen:
(a) The media stream overlap/gap may be introduced when the tRNS takes the
Serving RNS role and starts to produce the downlink data from forwarded
GTP-PDUs. In this case the estimated gap/overlap for hard handoff is equal
to the delay of the GTP tunnel used for data forwarding. This first instance of
media stream overlap coincides with radio hard handover.
(b) The additional media stream gap may be introduced when the CN transport is
optimized. In this case the gap will exist only if the delay via the optimized
route is larger than the delay via the forwarding route.
In comparison, for downlink media streams during the proposed handoff procedures,
there are only one possible situation when media stream gap (but no overlapping) may
happen. That is, the media stream overlap/gap may be introduced when the tRNS takes
the serving RNS role and starts to request a new set of MDSs for the rest of video
delivery. In this case the estimated gap for hard handoff is equal to the delay of stream re-
establishment for media delivery. This coincides with radio hard handoff.
If the transport bearer delay difference is smaller than the air interface Transmission
Time Interval (TTI) (10, 20, 40 or 80 ms depending on the service), the amount of gap is
most likely not existent.
In addition, as discussed previously, there is no buffer requirement in the BSs for
data forwarding and resequencing in case of stream re-establishig. Only a smaller buffer
is needed in the BSs for absorbing the delay jitter of a video stream and for re-ording due
to changes in routing paths. The relatively small queue in the BSs also reduces the
handoff latency.
74
2. It has relatively low packet loss (or frame error), and end-to-end delay.
Since the media streaming services are pushed to the edge of core network and the
streaming media can be delivered over a shorter network path, the transfer delay and
delay jitter of media service delivery, the probability of packet loss, and the total network
resource occupation will be reduced.
Furthermore, with the employment of stream re-establishig, the relatively small
queue in the BSs reduces the end-to-end delay further.
3. There is no extra handoff latency introduced due to session migration.
Since the technique of layered coding, e.g., SMDC, is employed as an alternative of
trancoding to combat the network heterogeneity or receiver heterogeneity, there are no
migration of transcoding state information required. Also, the migration of session
description is not necessary because of the principle of stream re-establishing instead of
data forwarding. Thus, only the migration of session parameters at the handoff decision
point is required. The amount of state information is thereby small enough to be hidden
inside the handoff signalling, so that there are no extra handoff latency introduced due to
session migration.
4. It has relatively consistent QoS in all scenarios (Handoff scalability
enhancement).
In UMTS Release 4, the values of handoff latency (i.e., delay jitter) vary with the
lengths of data- forwarding path in different handoff scenarios. Also, the end-to-end delay
varies with different delivery paths and different locations of the media providers, which
is outside the CN and far from the mobile hosts.
However, due to the introduction of the distributed multimedia delivery mobile
network, the values of handoff latency and end-to-end delay in different handoff
scenarios depend mainly on the length of media delivery path from MDs to SGSN/MSC,
and then to MS. Usually, the SGSN/MSC and the MDs are neighbor nodes. It is
straightforward that the length of media delivery path from the MDs to the MS is
relatively consistent in different handoff scenarios.
5. The amount of signalling traffic is slightly reduced during inter-RAN handoff.
A brief summary of above comparison between our proposal and that of UMTS is
given in Table 4-4.
75
The proposed procedure in the scenario of inter-cell, intra-RNS handoff can be
found in Appendix. Note that, in the scenario of intra-cell handoff, it is not necessary to
re-establish media stream and the corresponding handoff procedures in UMTS R99 may
be extended to support media streaming services.
Table 4-4 Summary of handoff solution comparison
UMTS D-MDMN
Principle Data forwarding Stream re-establishing
Queue size in the BSs Large Small
Packet loss (Frame error) High Low
End-to-end delay High Low
Handoff latency
(gap or overlapping)
High
(For downlink, two
instances of stream gap/
overlapping may occur)
Low
(For uplink, only one
instance of stream gap
may occur)
Consistency of QoS in all
scenarios
(i.e., Handoff scalability issue)
Poor Good
intra-RAN handoff 12
13 Signalling
traffic
inter-RAN handoff 18
16
This chapter presents the details of the proposed solutions for video mobility under
the system model defined in Chapter 3. In Section 4.1, the Scalable Multiple Description
Coding framework is proposed to explore the joint design of layered coding and multiple
description coding. Section 4.2 describes a novel IP DiffServ video marking algorithm to
support the UEP of SMDC, which re-organizes the shape, motion, and texture
information of video stream into different layers in order to implement the DiffServ in
UMTS. Finally, the corresponding intra-RAN handoff and inter-RAN handoff procedures
in D-MDMN are studied in Section 4.3 with the employment of the principle of video
stream re-establishing for seamless handoff.
76
5 Simulations
5.1 Simulation Models
The simulation model of UMTS intra-RAN and inter-RAN handoff are shown in
Figure 5-1 and Figure 5-3, respectively. Correspondingly, the simulation model of D-
MDMN intra-RAN and inter-RAN handoff are proposed as illustrated in Figure 5-2 and
Figure 5-4, respectively. The parameters and configuration attributes of the simulation
model can be chosen for different simulation scenarios.
The difference between UMTS and MDMN model is that the central Video Provider
outside the CN in the UMTS is distributed and pushed to the edge of the RAN in the
MDMN. For the sake of simplicity, one practical topology of multimedia delivery
networks in the real world is that each distributed media server is simplified as a
multimedia database into each SGSN. IP DiffServ is implemented in each node within
both UMTS and MDMN.
Note that since the bandwidth fluctuations and limitation of the wireless channels
are what we are more concerned with, we set up the system such that no congestion
happens at wired nodes, except for the BSs.
A brief description of the network node and link models and their roles is presented
below.
Radio Access Network
RAN is modeled as RNS and Wireless AP in our simulation. RNS node model is
shown in Figure 5-5. It consists of RNS (data plane), RNS (control plane), AN router and
IP-based AN. The radio functionality of Base Station (Node B) is implemented in the
Wireless AP. The data functionalities of BS and RNC are implemented in the RNS (data
plane); the control functionalities of BS and RNC are in the RNS (control plane). The
protocol stack is illustrated in Figure 3-7. The scale of the radio access network is
dependent on the attributes (e.g., packet latency and packet loss ratio) of IP-based AN
77
cloud model. In our simulation, the packet latency of IP-based AN is configured as
exponential distributed with mean value of 15 ms; the packet loss ratio is zero.
The BS acts as the extension of mobile clients. It handles the difference between
wireless and wired networks. Each BS is generally responsible for wireless connection
setup, handoff support, and medium access control in its service area. For streaming
video applications, it also has the responsibility of QoS control, such as rate filtering,
scheduling and ARQ, for media streaming.
The Packet Analyzer is used for OPNET ACE Tools to capture packet traces, which
will not affect the simulation results.
SGSN/GGSN and Video Provider
SGSN/GGSN node model in UMTS is shown in Figure 5-6. It is composed of
SGSN/GGSN (data plane), SGSN/GGSN (control plane) and Access Router. Figure 5-7
shows the SGSN/GGSN node model in MDMN, where Video Providers are distributed as
multimedia databases. The protocol stacks of SGSN/GGSN and Video Provider are
illustrated in Figure 3-7.
IP backbone core network
The scale of the CN is dependent on the attributes (e.g., packet latency and packet
loss ratio) of IP backbone CN cloud model. The packet latency of IP backbone CN is
configured as exponential distribution with mean value of 20 ms; the packet loss ratio is
zero.
Mobile Station
MS node model is shown in Figure 5-8, which is implemented according to the
protocol stack illustrated in Figure 3-7. The RFL, RNL, ip, ip_encap, tcp, udp, rtp and
application layer processors are taken from OPNET Modelers library. Each MS supports
three types of services: video streaming service, voice service and web service. The voice
and web services are configured as background traffic which are established between IP
Phone User and Voice User, Web Server and Web User, respectively.
78
Figure 5-1 UMTS simulation model (Intra-RAN Handoff)
Figure 5-2 MDMN simulation model (Intra-RAN Handoff)
79
Figure 5-3 UMTS simulation model (Inter-RAN Handoff)
Figure 5-4 MDMN simulation model (Inter-RAN Handoff)
80
Figure 5-5 Node Model of RNS
Figure 5-6 Node Model of SGSN (or GGSN)
Figure 5-7 Node Model of SGSN in the proposed MDMN
81
Figure 5-8 Node Model of MS
Wired Links
The PPP point-to-point links at OC3 data rate (155Mbps) are used between nodes
with serial interfaces (e.g., routers with PPP ports) within the RAN and CN. The video
provider, IP phone user or web server is connected to a mobile network access point (e.g.,
GGSN) by using a 100BaseT duplex link at 100 Mbps.
Radio Links
Unlike point-to-point links, radio links are not statically represented; that is, they
cannot be seen in the network model. Instead, radio links are dynamically established
during simulation. Radio links exist between any radio transmitter-receiver channel pair,
but establishing a link depends on many physical characteristics of the components
involved, as well as time-varying parameters. During simulation, parameters such as
frequency band, modulation type, transmitter power, distance, and antenna directionality
are common factors that determine whether a radio link exists at a particular time or can
ever exist.
OPNET uses the radio transceiver pipeline model, shown in Figure 5-9, to model
wireless transmission of packets. It consists of thirteen stages. The attributes of each stage
are configured as shown in Figure 5-10. The stage 10 to stage 13 in the radio pipeline
model which are related to wireless BER will be discussed in Section 5.2. For more
details of each stage in the radio pipeline model, please refer to the OPNET Modeler 9.0
documents [77].
82
Figure 5-9 Radio Transceiver Pipeline Model
Figure 5-10 Radio Transceiver Attributes for Specifying Pipeline Stages
83
5.2 System Setup and Test Conditions
5.2.1 Rate Shaping (Filtering) and Packetization
The purpose of rate shaper (filter) [1] [2] [3] [23] includes: optimization of
bandwidth usage, adoption of filter for handling client heterogeneity, network
heterogeneity, and optimizations in the retrieval of stored media.
Figure 5-11 Level of rate shaping
In Figure 5-11, the points at which rate shaping can be performed on the compressed
bit-stream are illustrated.
Region A is the uncompressed raw video where rate shaping can be preformed
relatively simply, e.g., resizing/stretching, but the amount of data that has to be processed
is very large due to its uncompressed nature.
In region B, the data is the same size as the raw data, but if the bit-stream is being
decompressed in order to accomplish a particular rate shaping and then recompressed (e.g.,
re-quantization filtering), performing filtering at this point saves completing the
computationally intensively functions of forward-DCT (FDCT) and inverse-DCT (IDCT).
In region C, the many zeros produced by the FDCT have been removed and the data
is considerably smaller, the functions of frequency filtering that are feasible at this point
84
include: low-pass filtering, color-reduction filtering, color to monochrome conversion
and simple coder-conversions. The current transcoder which is designed for speed
therefore operates in this region.
Compared with region A, B and C, rate shaping (e.g., frame dropping, layer dropping)
on the fully compressed data in region D is standard specific and relatively simple. We
employ the techniques of rate shaping in the sender directly to video distribution services as
network filtering used in the BSs. For simplicity of simulation, we packetize each frame of
every layer into one packet in order to mimic the principle of a frame dropper in the BSs.
5.2.2 BER over Rayleigh Fading Channels
As mentioned previous, the BER on wireless channels is computed at the BER stage
(stage 11) in the radio pipeline model. In general, the bit error rate provided by this stage
is a function of the type of modulation used for the transmitted signal. This stage
evaluates BER based on the previously computed average SNR and also accounts for
processing gain at the receiver.
The SNR (in dB) is given by
SNR = 10 log [ P
r
/
(N
b
+N
i
) ]
where P
r
= received power (Watts), which is computed in stage 7;
N
b
= background noise power (Watts), which is computed in stage 8;
N
i
= interference noise power (Watts), which is computed in stage 9.
The SNR value is added to the processing gain (also in dB) to obtain the effective
SNR. This effective SNR is also written as E
b
/ N
0
where E
b
is the received energy per bit
(in Joules) and N
0
is the noise power spectral density (in Watts/hertz). The bit error rate is
derived from the effective SNR based on the QPSK (downlink) / BPSK (uplink)
modulation curve assigned to the receiver in a Rayleigh fading channel. The probability
of bit error P
b
(i.e., BER) is given by [52]
( )
1
1
]
1
+
0
0
0
1
1
2
1
N E
N E
N E P
b
b
b b
Figure 5-12 shows the BER curves for QPSK (downlink) and BPSK (uplink) in a
Rayleigh fading channel and an AWGN channel.
85
Figure 5-12 BER performance over flat Rayleigh fading channels
Following the BER stage, the error allocation stage (stage 12) translates the given
BER computed in stage 11 into an actual set of bit errors for each valid packet which is
received. The approach taken in stage 12 is to avoid sequencing through all the bits in the
packet and, therefore, not to generate positional information about bit errors, but to still
accurately compute the overall number of bit errors. This should be done with the
minimum number of computations for simulation efficiency considerations.
The error allocation algorithm is based on the expression for the probability of
exactly k bit errors occurring in a packet segment of length N. This probability is denoted
as P
k
. Under the assumption of independent bit errors, given the wireless channel bit error
probability P
b
(i.e., BER), P
k
can be expressed by
r P
k
N
P P P
N
k
k
k N
b
k
b k
,
_
0
) 1 (
The algorithm first generates a random number r between 0 and 1, to provide a value
against which the probability of occurrences of different numbers of bit errors will be
86
tested. Next the algorithm begins iterations. First the probability that 0 bit error occurs is
computed according to the formula above, and compared to r. If r is lower than this
probability, then 0 is the number of bit errors in the packet. Else, the probability of
occurrence of 1 or fewer bit errors is computed. If r is less than this probability, yet
higher than the probability of occurrence of the previous number of bit errors, then 1 is
the number of bit errors allocated to the packet. The algorithm continues iterating in this
manner until a value k is found for which the probability of k or fewer errors occurring is
greater than r. Then the number of errors assigned to the packet is k.
The purpose of error collection stage (stage 13) is to determine whether or not the
arriving packet can be accepted in the destination node. This is usually dependent upon
whether the packet has experienced collisions, the result computed in the error allocation
stage, and the capability of the receiver to correct the errors affecting the packet. In our
simulation, there are no error collection techniques implemented except for ARQ.
The following three cases in Table 5-1 are tested for the evaluation of the
relationship between BER and frame error rate.
Table 5-1 Test Cases of BER and FER in wireless channels
Test Case Channel Model Modulation Multiple Access BER
Rayleigh Fading QPSK (DL), BPSK(UL) DS-CDMA [10
-5
, 2 10
-5
]
Rayleigh Fading QPSK (DL), BPSK(UL) DS-CDMA [10
-4
, 2 10
-4
]
Rayleigh Fading QPSK (DL), BPSK(UL) DS-CDMA [.6 10
-3
, 10
-3
]
5.2.3 Delay-constrained ARQ
To enhance the video quality in the presence of unavoidable packet loss or bit errors,
error control mechanisms have been proposed. Basically, error control approaches can be
broadly categorized as open- loop error control (e.g., SMDC, error resilience tools in
MPEG-4, error concealment) and close- loop error control, (e.g., delay-constrained ARQ).
However, because of the complexity of open- loop error control, multiple description
coding, error resilience tools in MPEG-4 and error concealment can not be emulated in
this simulation. Instead, delay-constrained ARQ is under consideration.
87
According to the decision- making points, delay-constrained retransmission can be
sender-based, receiver-driven, or hybrid. Since the computational complexity is limited at
the mobile handset, the sender-based control at the BS is chosen to suppress
retransmission of packets that will miss their display time at the mobile user.
Given the maximum transfer delay D
max
in UMTS, the wired link delay D
wired
and
the wireless link RTT
wireless
, the maximum number of retransmissions N
ARQ
allowed for a
video packet is approximately given by
'
>
1
]
1
wired
wired
wireless
wired
ARQ
D D if
D D if
RTT
D D
N
max
max
max
, 0
,
Typically, the wireless link RTT
wireless
is about several milliseconds. The BS can
calculates the wired link delay D
wired
from Video Provider to the BS by recording the
time T when a RTCP sender report (SR) is received, and then subtracting the value of
NTP timestamp field in SR to obtain D
wired
= (T - NTP).
5.2.4 Traffic Profile
In order to evaluate the effectiveness of the proposed IP DiffServ MPEG-4 video
marking algorithm and the streaming video handoff procedures under MDMN, two
groups of traffic parameters have been set up and are listed in Table 5-2 and Table 5-4,
respectively.
Traffic Profile for Evaluation of Video Marking Algorithm
The video traffic, shown in Figure 5-14, is made up of Class I, Class II and Class III
layer-coded video streams, shown in Figure 5-13, which are classified through the
proposed IP DiffServ MPEG-4 video marking algorithm. The video traffic which is
generated by OPNET has to be subjected to the requirements of UMTS bearer service
attributes of streaming class [61], listed in Table 5-3.
The background traffic in each cell is composed of one video client, 25 voice users,
and web users, shown in Figure 5-15 ~ Figure 5-17, respectively. Under the condition of
background traffic, the serious congestion will occur in the BSs due to the limitation and
fluctuation of wireless bandwidth.
88
Table 5-2 Traffic profile of each cell for evaluation of video marking algorithm
Traffic Type User Number Sending Rate (mean) Packet Length (mean) Standard
Video Stream* 2 240 kbps per client 1 Kbytes MPEG-4
Voice Stream 25 16 kbps per user 200 Bytes G.728
Web Traffic - 400 kbps 1 Kbytes HTTP
* Frame Rate (mean) = 20 frame/s, Frame Length (mean) = 1 Kbytes.
Table 5-3 UMTS bearer service attributes of streaming class
Maximum bit rate (outdoor) 384 kbps
Maximum SDU size 1500 Bytes
SDU error ratio (i.e., Frame error rate) 10
%
Transfer delay 280 ms
Delay jitter 50 ms (Frame Rate = 20 frame/s)
Traffic Profile for Evaluation of Streaming Video Handoff
Table 5-4 Traffic profile of each cell for evaluation of streaming video handoff
Traffic Type User Number Sending Rate (mean) Packet Length (mean) Standard
Video Stream* 2 240 kbps per client 1 Kbytes MPEG-4
Voice Stream 5 16 kbps per user 200 Bytes G.728
Web Traffic - 400 kbps 1 Kbytes HTTP
* Frame Rate (mean) = 20 frame/s, Frame Length (mean) = 1 Kbytes.
The video traffic, shown in Figure 5-18, is also made up of Class I, Class II and
Class III layer-coded video streams and has to be subjected to the requirements of UMTS
bearer service attributes of streaming class [61], listed in Table 5-3. The hard handoff will
occur 28 times during the testing period of 5 minutes.
The background traffic in each cell is composed of one video client, 5 voice users,
and web users, shown in Figure 5-19 ~ Figure 5-21, respectively. With the decrease of
background traffic, no congestion happens in the BSs during the handoff tests.
89
Figure 5-13 Layered Video Traffic (Video Client1)
Figure 5-14 Video Traffic (Video Client1)
90
Figure 5-15 Background Traffic (Video Client2)
Figure 5-16 Background Traffic (25 Voice Users)
Figure 5-17 Background Traffic (Web Users)
91
Figure 5-18 Video Traffic (Handoff occurs 28 times)
Figure 5-19 Background Traffic (Video Client s)
Figure 5-20 Background Traffic (5 Voice Users)
Figure 5-21 Background Traffic (Web Users)
92
5.2.5 Test Cases
AF Queue Size
To implement the proposed IP DiffServ video marking algorithm in each node, we
use WFQ discipline to classify and schedule the incoming packet into and out of EF,
AF1, AF2, AF3 or BE queue. The WFQ profile for the proposed IP DiffServ video
marking algorithm in each node is listed in Table 5-5. The selection of buffer size is
crucial to video over IP network. An optimum buffer size has to be found which balances
both end-to-end delay and packet loss ratio to tolerable levels. If the buffer is set too low,
some packets may be lost; if set too high, higher delays result.
Table 5-5 WFQ profile for proposed IP DiffServ video marking algorithm
Queue Scheduling
Queue
Classification
Queue Size
(1 unit = 100 ms)
Normalized
Bandwidth
BE Queue 15 4.5 %
AF3 Queue 1~9 9.1 %
AF2 Queue 1~9 13.7 %
AF1 Queue 1~9 22.7 %
WFQ
EF Queue 15 50 %
There are eighteen cases, listed in Table 5-6, under test in order to evaluate the effect
of AF queue size. The effect of AF queue size is analyzed in Section 5.2.5. After the
evaluation of the effect of AF queue size, we select 500 ms as the optimum queue size of
the AF1, AF2 and AF3 queue.
Video Marking Algorithm
In order to compare the proposed IP DiffServ video marking algorithm with DVMA,
three scenarios are experimented. The first one is the best effort model using a drop-tail
BE queue in each node. The second one is the DiffServ model where DVMA is used with
WRED AF queue. The third one is the DiffServ model where the proposed IP DiffServ
video marking algorithm is used with the drop-tail AF1, AF2 and AF3 queues. Table 5-7
lists the configuration of queue scheduling in these three scenarios. Table 5-8 gives the
93
WRED profile for AF Queue in the DVMA scenario. The test cases of video marking
algorithm are listed in Table 5-9.
Streaming Video Handoff
The performance of the proposed streaming video handoff in D-MDMN is examined
and compared with that in UMTS model under the scenario of the proposed IP DiffServ
video marking algorithm. The test cases of streaming video handoff are listed in Table 5-10.
Table 5-6 Test Cases of Effect of Queue Size
Test Case Physical Characteristics Rayleigh BER QoS Mechanism AF Queue Size
1 QPFK(DL),DS-CDMA 10
-5
DiffServ, WFQ 100 ms
2 QPFK(DL),DS-CDMA 10
-5
DiffServ, WFQ 200 ms
3 QPFK(DL),DS-CDMA 10
-5
DiffServ, WFQ 300 ms
4 QPFK(DL),DS-CDMA 10
-5
DiffServ, WFQ 400 ms
5 QPFK(DL),DS-CDMA 10
-5
DiffServ, WFQ 500 ms
6 QPFK(DL),DS-CDMA 10
-5
DiffServ, WFQ 600 ms
7 QPFK(DL),DS-CDMA 10
-5
DiffServ, WFQ 700 ms
8 QPFK(DL),DS-CDMA 10
-5
DiffServ, WFQ 800 ms
9 QPFK(DL),DS-CDMA 10
-5
DiffServ, WFQ 900 ms
(1) QPFK(DL),DS-CDMA 10
-4
DiffServ, WFQ 100 ms
(2) QPFK(DL),DS-CDMA 10
-4
DiffServ, WFQ 200 ms
(3) QPFK(DL),DS-CDMA 10
-4
DiffServ, WFQ 300 ms
(4) QPFK(DL),DS-CDMA 10
-4
DiffServ, WFQ 400 ms
(5) QPFK(DL),DS-CDMA 10
-4
DiffServ, WFQ 500 ms
(6) QPFK(DL),DS-CDMA 10
-4
DiffServ, WFQ 600 ms
(7) QPFK(DL),DS-CDMA 10
-4
DiffServ, WFQ 700 ms
(8) QPFK(DL),DS-CDMA 10
-4
DiffServ, WFQ 800 ms
(9) QPFK(DL),DS-CDMA 10
-4
DiffServ, WFQ 900 ms
94
Table 5-7 Queue scheduling configuration
Solution
Queue
Scheduling
Queue
Classification
Queue Size
(1 unit = 100ms)
Normalized
Bandwidth
Best Effort FIFO BE Queue 45 100 %
BE Queue 15 4.5 %
AF Queue 15 45.5 %
DVMA WRED+CBQ
EF Queue 15 50 %
BE Queue 15 4.5 %
AF3 Queue 5 9.1 %
AF2 Queue 5 13.7 %
AF1 Queue 5 22.7 %
Proposed WFQ
EF Queue 15 50 %
Table 5-8 WRED profile for AF queue in DVMA
Video Stream
Class
Exponential
Weight Factor
Min Threshold
(1 unit = 100 ms)
Max Threshold
(1 unit = 100 ms)
Mark Probability
Denominator
AF31 9 1 5 5
AF21 9 2 5 5
AF11 9 3 5 5
Table 5-9 Test Cases of Video Marking Algorithm
Test Case Physical Characteristics
Rayleigh
BER
QoS
Mechanism
Queue
Scheduling
Marking
Algorithm
A QPFK(DL), DS-CDMA 10
-5
Best Effort IP FIFO BE
B QPFK(DL), DS-CDMA 10
-5
IP DiffServ WRED DVMA
C QPFK(DL), DS-CDMA 10
-5
IP DiffServ WFQ Proposed
a QPFK(DL), DS-CDMA 10
-4
Best Effort IP FIFO BE
b QPFK(DL), DS-CDMA 10
-4
IP DiffServ WRED DVMA
c QPFK(DL), DS-CDMA 10
-4
IP DiffServ WFQ Proposed
95
Table 5-10 Test Cases of Handoff Procedures
Test
Case
Modulation
Rayleigh
BER
Network
Model
QoS Mechanism Handoff
I QPSK(DL), BPSK(UL) 10
-5
UMTS DiffServ, WFQ Intra-RAN
II QPSK(DL), BPSK(UL) 10
-5
MDMN DiffServ, WFQ Intra-RAN
III QPSK(DL), BPSK(UL) 10
-5
UMTS DiffServ, WFQ Inter-RAN
IV QPSK(DL), BPSK(UL) 10
-5
MDMN DiffServ, WFQ Inter-RAN
i QPSK(DL), BPSK(UL)
10
-4
UMTS DiffServ, WFQ Intra-RAN
ii QPSK(DL), BPSK(UL)
10
-4
MDMN DiffServ, WFQ Intra-RAN
iii QPSK(DL), BPSK(UL)
10
-4
UMTS DiffServ, WFQ Inter-RAN
iv QPSK(DL), BPSK(UL)
10
-4
MDMN DiffServ, WFQ Inter-RAN
5.3 Simulation Results and Analysis
5.3.1 Effect of BER on FER over Wireless Channels
Figure 5-22 ~ Figure 5-27 show the effect of BER on FER over the Rayleigh fading
channel. The FER measure indicates the difficulty of sending video stream over the
wireless channel. A small BER translates into a much higher FER. For example, the BER
mean value of 1.3610
-4
in Figure 5-23 can translate into the FER mean value of 3.16%
in Figure 5-26. In the test case with a BER mean value of 1.5710
-5
, there are no
frame errors showed in Figure 5-25. That is because we use the delay-constrained ARQ
to combat Rayleigh BER.
Given that the maximum frame error rate allowed in UMTS is 10%, the video delivery
in UMTS can tolerate Rayleigh BER up to 10
-4
with delay-constrained ARQ. That is reason
that we choose BER in the range of [10
-5
, 2 10
-5
]
and [10
-4
, 2 10
-4
] as our test conditions.
However, from Table 1-1, the BER of wireless video can be as high as 10
-3
. It is
desirable to employ the proposed scalable multiple description coding as a diversity
technique to combat the high BER in the wireless channel.
96
Figure 5-22 Bit Error Rate over the Rayleigh fading channel (Test Case )
Figure 5-23 Bit Error Rate over the Rayleigh fading channel (Test Case )
Figure 5-24 Bit Error Rate over the Rayleigh fading channel (Test Case )
97
Figure 5-25 FER over the Rayleigh fading channel (Test Case )
Figure 5-26 FER over the Rayleigh fading channel (Test Case )
Figure 5-27 FER over the Rayleigh fading channel (Test Case )
98
5.3.2 Effect of AF Queue Size
Figure 5-28 ~ Figure 5-31 illustrate the effect of the AF queue size on packet end-to-
end delay and packet loss ratio of AF queue at the BSs. The video traffic begins at time
25 s and the background traffic starts at time 36 sec. From time 25 s to 36 s, the packet
arriving rate at BS is lower than the queue service rate and there is no congestion in the
AF queues. At time 36 s, the packet arriving rate at BS suddenly increases due to the
background traffic and exceeds the queue service rate. The AF queues continue growing
until time 50 s and start to drop frames.
As the AF queue size increases, the packet end-to-end delay increases, but the
packet loss ratio of AF queue decreases. Compared the results of total nine test cases, we
choose queue size 500 ms as the optimum value which balances both end-to-end delay
and packet loss ratio to tolerable levels for further simulation setting.
5.3.3 Effect of Video Marking Algorithm
Performance of Frame Error Rate
MPEG-4 video stream in UMTS can tolerance SDU error rate (i.e., FER) up to 10%.
The effects of different video marking algorithms on FER are shown in Figure 5-32 ~
Figure 5-37.
In the scenario of BER = 10
-5
, the FERs are caused by the packet loss at the BSs due
to congestion. In the scenario of BER = 10
-4
, the FERs are caused by both the packet loss
at the BSs due to congestion and the wireless channel bit errors. Due to the employment
of WRED for proactive packet-dropping in DVMA, the packet loss in DVMA begins
earlier than that in the proposed solution. However, the packet loss in the tail-dropping
BE solution occurrs even earlier than that in DVMA. That is because the background
traffic and the video traffic enter into the same queue, and cause the congestion happened
earlier. Note that the background traffic and the video traffic are separated in different
queues in DiffServ-based solutions.
Since we re-organize the shape, motion, and texture video information into different
layers, UEP can be introduced and results in differentiated service. If a bit error or a
99
packet loss occurs in the MPEG-4 Class I stream, the corresponding bits or packets in the
MPEG-4 Class II and Class III streams have to be considered erroneous or lost. Similarly,
if a bit error or a packet loss occurs in the MPEG-4 Class II stream, the corresponding
bits or packets in the MPEG-4 Class III stream also have to be considered erroneous or
lost. Some packets may arrive late and will also be considered lost. If the higher priority
traffic is protected, less packet loss (i.e., FER) will occur.
Compared with BE and DVMA, the protection of both voice stream and MPEG-4
Class I stream in the proposed solution is the best. This also results in the least FER
among all the scenarios. In addition, the protection of EF traffic in DVMA is better than
that in BE. However, this is at the cost of high FER of AF traffic in DVMA.
A brief comparison of QoS (e.g., FER) guarantee in the three solutions can be
summarized in Table 5-11. Note that: Unacceptable QoS means that the performance of
FER can not meet the QoS requirement all the time; Unpredictable QoS is a typical
characteristic of best-effort traffic; Unwarranted QoS means that the performance of
FER meets the QoS requirement sometimes and is predictable; Guaranteed QoS means
that the performance of FER meets the QoS requirement all the time and is predictable.
Table 5-11 Comparison of QoS (e.g., FER) guarantee in three solutions
Traffic BE DVMA Proposed
Voice Stream Unacceptable QoS Unwarranted QoS Guaranteed QoS
MPEG-4 Class I Unpredictable QoS Unwarranted QoS Guaranteed QoS
MPEG-4 Class II Unpredictable QoS Unwarranted QoS Unwarranted QoS
MPEG-4 Class III Unpredictable QoS Unwarranted QoS Unwarranted QoS
Performance of End-to-end Delay and Delay Jitter
The maximum end-to-end delay and delay jitter of video streaming allowed in the
simulation are 280 ms and 50 ms, respectively, under the test conditions. The effects of
different video marking algorithms on end-to-end delay and delay jitter are presented in
Figure 5-38 ~ Figure 5-49.
In the BE solution and the DVMA solution, the end-to-end delay and delay jitter of
video traffic in different classes can not be differentiated. That is because different
100
classes video streams go through the same queue (e.g., BE queue or AF queue). As we
expect in the proposed solution, the performance of MPEG-4 Class I stream is better than
Class II; and Class II is better than Class III.
In the scenario of BER = 10
-4
, with a sudden increase of the background traffic at
time 36 s, the end-to-end delay and delay jitter of video streams jump sharply up to a
higher level. The proposed solution delays 10 s the start point of performance degradation
compared with BE and DVMA. After the BSs begin to drop packets, the performance of
end-to-end delay and delay jitter turns better. In DVMA, the video delay jitters of all
three classes are not acceptable, though Class III and Class II streams in DMVA are
better than those in the proposed solution. In comparison, the MPEG-4 Class III and
Class II stream in the proposed solution are sacrificed in order to guarantee the QoS of
the Class I stream.
The performance tendency is similar in the scenario of BER = 10
-5
. Note that only
MPEG-4 Class III stream in the proposed solution suffers from high delay and delay jitter
in order to preserve the Class I and Class II streams. However, in DVMA, all three
classes are not acceptable until about time 1m40s. The results in the scenario of BER =
10
-5
is better than their counterparts in the scenario of BER = 10
-4
. This is because the
numbers of ARQ in BER = 10
-5
is less than that in BER = 10
-4
.
In both BER scenarios, the performance of delay and delay jitter in BE seems better
than the other two solutions. However, this is at the cost of unacceptable FER in the voice
stream and unpredictable quality (e.g., FER) of streaming video service.
5.3.4 Effect of Streaming Video Handoff
As discussed previously, the maximum end-to-end delay and delay jitter of video
streaming allowed in both UMTS and MDMN are 280 ms and 50 ms, respectively, under
the test conditions. The handoff latency also should keep below 50 ms as a delay jitter.
Figure 5-50 ~ Figure 5-57 show the performance of end-to-end delay and delay jitter
in the scenario of BER = 10
-5
. Figure 5-58 ~ Figure 5-65 illustrate the performance of
end-to-end delay and delay jitter in the scenario of BER = 10
-4
.
In UMTS, only the performance of maximum end-to-end delay in test case iii (Inter-
RAN Handoff, BER = 10
-4
) exceeds 280 ms. Furthermore, in all the UMTS test cases, the
101
handoff latency can not satisfy the 50 ms delay bound. Note that, because the distance
between the Video Provider and GGSN is unknown, we choose the best case in the
UMTS simulation model. That is, the distance between them is only one hop. If the
Video Provider is far from the UMTS core network, the performance of maximum end-
to-end delay in UMTS may not meet the QoS requirement of UMTS any more.
In comparison, with the proposed stream re-establishing handoff solution, all the test
cases in MDMN meet the performance requirement of the maximum end-to-end delay
and handoff latency. Because the Video Provider is distributed as the media databases
inside the core network, the handoff performance keeps stable in MDMN and does not
have the distance problem as mentioned above in UMTS.
A brief comparison of handoff performance in UMTS and MDMN is summarized in
Table 5-12. This comparison shows that the improvement of handoff performance
ascends with the increase of the scale of the mobile core network. For example, the
handoff latency improvement in the intra-RAN scale is 26 ms or 45 ms under different
conditions of wireless BER, but in the inter-RAN scale it is 38 ms or 57 ms. Furthermore,
the proposed stream re-establishing handoff performance in MDMN is relatively
consistent in all scenarios. This validates the enhancement of handoff scalability in
MDMN.
Table 5-12 Comparison of handoff performance in UMTS and MDMN
Test Case
End-to-end Delay
(mean)
Delay
Handoff Latency
(mean)
Handoff
Latency
I (UMTS, Intra-RAN) 74 ms 68 ms
II (MDMN, Intra-RAN) 42 ms
32 ms
42 ms
26 ms
III (UMTS, Inter-RAN) 75 ms 91 ms
IV (MDMN, Inter-RAN) 45 ms
30 ms
46 ms
45 ms
i (UMTS, Intra-RAN) 79 ms 82 ms
ii (MDMN, Intra-RAN) 45 ms
34 ms
44 ms
38 ms
iii (UMTS, Inter-RAN) 131 ms 104 ms
iv (MDMN, Inter-RAN) 56 ms
75 ms
47 ms
57 ms
102
In this chapter, the simulation models, system setup, test conditions, and simulation
results are presented and analyzed. Section 5.1 describes the UMTS and the proposed D-
MDMN simulation model, and the simulation pipeline stage of the radio transceiver. In
Section 5.2, we discuss the simulation and design issues, such as rate shaping and
packetization, BER over Rayleigh fading channels, delay-constrained ARQ, followed by
the description of traffic profile and test cases. Section 5.2.5 first evaluates the effect of
BER on FER over wireless channels and the effect of AF queue size for optimization of
system setup. Then performance evaluation of the proposed IP DiffServ video marking
algorithm is undertaken to show that it is more suitable for video streaming in IP mobile
networks compared with DVMA solution. Finally, the simulation analysis is concluded
by the handoff performance comparison of UMTS versus D-MDMN, indicating that the
proposed handoff procedures in D-MDMN have better performance in terms of handoff
latency, end-to-end delay and handoff scalability than that in UMTS.
103
Figure 5-28 Average of End-to-End Delay (Test Case 1~9: BER = 10
-5
)
Figure 5-29 Average of Packet Loss Ratio (Test Case 1~9: BER = 10
-5
)
104
Figure 5-30 Average of End-to-End Delay (Test Case (1)~(9): BER = 10
-4
)
Figure 5-31 Average of Packet Loss Ratio (Test Case (1)~(9): BER = 10
-4
)
105
Figure 5-32 FER (Test Case A: Best Effort, FIFO, BER = 10
-5
)
Figure 5-33 FER (Test Case B: DiffServ, WRED, BER = 10
-5
)
106
Figure 5-34 FER (Test Case C: DiffServ, WFQ, BER = 10
-5
)
Figure 5-35 FER (Test Case a: Best Effort, FIFO, BER = 10
-4
)
107
Figure 5-36 FER (Test Case b: DiffServ, WRED, BER = 10
-4
)
Figure 5-37 FER (Test Case c: DiffServ, WFQ, BER = 10
-4
)
108
Figure 5-38 End-to-end Delay (Test Case A: Best Effort, FIFO, BER = 10
-5
)
Figure 5-39 End-to-end Delay (Test Case B: DiffServ, WRED, BER = 10
-5
)
Figure 5-40 End-to-end Delay (Test Case C: DiffServ, WFQ, BER = 10
-5
)
109
Figure 5-41 End-to-end Delay (Test Case a: Best Effort, FIFO, BER = 10
-4
)
Figure 5-42 End-to-end Delay (Test Case b: DiffServ, WRED, BER = 10
-4
)
Figure 5-43 End-to-end Delay (Test Case c: DiffServ, WFQ, BER = 10
-4
)
110
Figure 5-44 Delay Jitter (Test Case A: Best Effort, FIFO, BER = 10
-5
)
Figure 5-45 Delay Jitter (Test Case B: DiffServ, WRED, BER = 10
-5
)
Figure 5-46 Delay Jitter (Test Case C: DiffServ, WFQ, BER = 10
-5
)
111
Figure 5-47 Delay Jitter (Test Case a: Best Effort, FIFO, BER = 10
-4
)
Figure 5-48 Delay Jitter (Test Case b: DiffServ, WRED, BER = 10
-4
)
Figure 5-49 Delay Jitter (Test Case c: DiffServ, WFQ, BER = 10
-4
)
112
Figure 5-50 End-to-End Delay (Test Case I: UMTS, Intra-RAN, BER = 10
-5
)
Figure 5-51 Delay Jitter (Test Case I: UMTS, Intra-RAN, BER = 10
-5
)
113
Figure 5-52 End-to-End Delay (Test Case II: MDMN, Intra-RAN, BER = 10
-5
)
Figure 5-53 Delay Jitter (Test Case II: MDMN, Intra-RAN, BER = 10
-5
)
114
Figure 5-54 End-to-End Delay (Test Case III: UMTS, Inter-RAN, BER = 10
-5
)
Figure 5-55 Delay Jitter (Test Case III: UMTS, Inter-RAN, BER = 10
-5
)
115
Figure 5-56 End-to-End Delay (Test Case IV: MDMN, Inter-RAN, BER = 10
-5
)
Figure 5-57 Delay Jitter (Test Case IV: MDMN, Inter-RAN, BER = 10
-5
)
116
Figure 5-58 End-to-End Delay (Test Case i: UMTS, Intra-RAN, BER = 10
-4
)
Figure 5-59 Delay Jitter (Test Case i: UMTS, Intra-RAN, BER = 10
-4
)
117
Figure 5-60 End-to-End Delay (Test Case ii: MDMN, Intra-RAN, BER = 10
-4
)
Figure 5-61 Delay Jitter (Test Case ii: MDMN, Intra-RAN, BER = 10
-4
)
118
Figure 5-62 End-to-End Delay (Test Case iii: UMTS, Inter-RAN, BER = 10
-4
)
Figure 5-63 Delay Jitter (Test Case iii: UMTS, Inter-RAN, BER = 10
-4
)
119
Figure 5-64 End-to-End Delay (Test Case iv: MDMN, Inter-RAN, BER = 10
-4
)
Figure 5-65 Delay Jitter (Test Case iv: MDMN, Inter-RAN, BER = 10
-4
)
120
6 Conclusions and Future Work
To address the handoff problems in video streaming, as well as the bandwidth
fluctuation, packet loss and heterogeneity problems in the wireless networks, and to
further enhance the error resilience tools in MPEG-4, the 3G mobile network architecture
and the handoff procedures for video delivery in UMTS are studied.
The contributions of this thesis are:
1) A Scalable Multiple Description Coding framework is proposed to explore the
joint design of layered coding and multiple description coding.
Under the SMDC framework, MDC components enhance the robustness to losses
and bit errors of LC components through path diversity and error recovery. MDC
components also reduce the storage, reliability and load balancing requirement among
distributed media edge servers. At the same time, LC components not only deal with the
unbalanced MD operation at the server end, but also combat the bandwidth frustrations of
the time- varying wireless channel. Furthermore, SMDC leverages the distributed
multimedia delivery mobile network to provide path diversity to combat video streaming
outage due to handoff.
2) A Distributed Multimedia Delivery Mobile Network is proposed for the UMTS
core network.
D-MDMN introduces and combines the concepts of CDN and SMDC into the
UMTS network in order to solve the video handoff problem and meet the stringent QoS
requirements of video streaming in 3GPP. Since the media streaming services are pushed
to the edge of core network, it also reduces the media service delivery time, the
probability of packet loss, and the total network resource occupation with relatively
consistent QoS in all scenarios.
3) Handoff procedures of video streaming in D-MDMN are proposed.
The proposed handoff procedures employ the principle of video stream re-
establishing to replace the principle of data forwarding in UMTS. The intra-RAN handoff
and inter-RAN handoff procedures are studied in details.
121
4) A novel IP DiffServ video marking algorithm is proposed to support the unequal
error protection of LC components of SMDC.
The proposed algorithm re-organizes the shape, motion, and texture information of
video stream into different layers in the proposed SMDC scheme to implement the
DiffServ mobile network in UMTS. Furthermore, it spurs the evolution of UMTS toward
its final all-IP phase for the purpose of addressing the DiffServ tunneling issue in UMTS.
The above proposed schemes have been validated through the simulation presented
in Chapter 5, except that the verification of MDC components of SMDC can not be
undertaken because of technical complexity and time limitation.
The limitations and cost of the proposed schemes also can be summarized as
follows.
1) The significant performance improvement of SMDC is achieved at the cost of a
coding overhead.
2) The D-MDMN network solution is feasible as a client-server solution for video
streaming delivery service, but not for end-to-end real- time video conversation.
3) For video streaming delivery, the video descriptions should be distributed in
advance into all complementary MDSs at the edge of the RANs according to the service
subscription of video mobile users, which adds the complexity to UMTS. A potential
video adaptive deployment solution can be given as follow.
Firstly, during the CAC, the video service subscribed by an MS are distributed to the
MDSs in the current local RAN and all its neighbor RANs. Secondly, if the MS moves to
another RAN, this video service should be simultaneously distributed to the MDSs in all
its new neighbor RANs after the handoff successfully takes place.
Further work will include the verification of the proposed SMDC under the object-
based MPEG-4 video stream, especially the MDC components over the Rayleigh fading
channel, the study of the effect of SMDC coding overhead and the soft handoff
procedures in D-MDMN. In addition, the proposed solutions of video streaming may be
applicable to audio streaming. The joint design of audio LC and MDC, the distributed
audio delivery network, the IP DiffServ audio marking algorithm, and the media
synchronization between video stream and audio stream should be studied further.
122
Appendix
Proposed inter-cell, intra-RNS handoff procedure
In the scenario of inter-cell, intra-RNS handoff, the proposed handoff procedures,
shown in Figure 0-1, consists of three phase:
l Phase I: Preparation of BTS handoff and resource allocation
l Phase II: Moving the Serving BTS role to target BTS
l Phase III: Releasing resource reservation in the old path
Figure 0-1 Proposed inter-cell, intra-RNS handoff procedure (Control plane)
123
Acronyms
3GPP Third Generation Partnership Project
ABR Available Bit Rate
AF Assured Forwarding
ALP Application Level Packets
AMR Adaptive Multi Rate
AR Access Router
ARQ Automatic Repeat reQuest
ATM Asynchronous Transfer Mode
BE Best Effort
BER Bit Error Rate
B-frame Bi-directionally predicted frame
BIFS Binary Format for Scene
BPSK Binary Phase Shift Keying
BR Border Router
BS Base Station
BSC Base Station Controller
BSS Base Station Subsystem
BTS Base Transceiver Station
CAC Call Admission Control
CBQ Class Based Queuing
CBR Constant Bit Rate
CCIR Consultative Committee for International Radiocommunication
CDMA Code Division Multiple Access
CDN Content Delivery Network
CIF Common Interleaved Frame
CLR Cell Loss Rate
CN Core Network
124
CoS Class of Service
CS Circuit Switched
DC Direct Current
DCT Discrete Cosine Transform
DiffServ Differentiated Services
DL Down Link
D-MDMN Distributed Multimedia Delivery Mobile Network
DS DiffServ
DSCP DiffServ CodePoint
DVMA DiffServ video marking algorithm
EDGE Enhanced Data rates for GSM Evolution
EF Expedited Forwarding
ER Edge Router
FDCT Forward-DCT
FDD Frequency Division Duplex
FEC Forward Error Correction
FER Frame Erasure Rates
FGS Fine granularity scalability
FP Frame relay transport Protocol
GERAN GSM/EDGE radio access network
GFR Guaranteed Frame Rate
GGSN Gateway GPRS Support Node
GOB Group Of Blocks
GOP Group Of Pictures
GPRS General Packet Radio Service
GSM Global System for Mobile communications
GTP GPRS Tunneling Protocol
HLR Home Location Register
IDCT Inverse-DCT
IEEE Institute of Electrical and Electronics Engineers
IETF Internet Engineering Task Force
125
I-frame Intra-coded frame
IP Internet Protocol
ISDN Integrated Services Digital Network
ITU International Telecommunication Union
I-VOP Intra-coded Video Object Plane
LAN Local Area Network
LC Layered Coding
MC-CMDA Multiple Carrier CDMA
MD Multiple Description
MDC Multiple Description Coding
MDS Media Description Server
MPEG Moving Pictures Experts Group
MS Mobile Station
MSC Mobile Switching Center
nrt-VBR non real-time Variable Bit Rate
NSS Network Subsystem
OD Object Descriptor
OFDM Orthogonal Frequency Division Multiplexing
PDU Protocol Data Unit
PFGS Progressive Fine Granularity Scalability
PFGST PFGS Temporal
P-frame Predictively coded frame
PHB Per-Hop Behavior
PS Packet Switched
PSS Packet-switched Streaming Service
PSTN Public Switched Telephone Network
P-VOP Predictively coded Video Object Plane
QCIF Quarter Common Interleaved Frame
QoS Quality of Service
QPSK Quadrature Phase Shift Keying
RAN Radio Access Network
126
RED Random Early Detection
RFC Request For Comments
RFL Radio Frequency Layer
RNC Radio Network Controller
RNL Radio Network Layer
RNS Radio Network Subsystems
RPS Reference Picture Selection
RS Redirection Server
RTCP Real Time Control Protocol
RTP Real Time Protocol
RTSP Real Time Streaming Protocol
RTT Round Trip Time
SD Single Description
SDP Session Description Protocol
SDU Service Data Unit
SGSN Serving GPRS Support Node
SIP Initiation Protocol
SLA Service Level Agreement
SMDC Scalable Multiple Description Coding
SNR Signal- to-Noise Ratio
sRNS source RNS
sSGSN source SGSN
TCP Transmission Control Protocol
TD-CDMA Time Division-CDMA
TDD Time Division Duplex
tRNS target RNS
tSGSN target SGSN
TTI Transmission Time Interval
UDP User Datagram Protocol
UE User Equipment
UEP Unequal Error Protection
127
UL Up Link
UMTS Universal Mobile Telecommunications System
URL Uniform Resource Locator
UTRAN UMTS Terrestrial Radio Access Network
VLR Visitor Location Register
VO Video Object
VoIP Voice over IP
VOP Video Object Planes
VP Video Packet
WAN Wide Area Network
WCDMA Wideband-CDMA
WFQ Weighted Fair Queuing
WRED Weighted Random Early Detection
128
References
[1] Dapeng Wu, Y. T. Hou, Wenwu Zhu, Ya-Qin Zhang, Jon M. Peha, Streaming
Video over the Internet: Approaches and Directions, IEEE Trans. on Circuits
and Systems for Video Technology, Vol. 11 No. 3 , pp. 282 -300, Mar. 2001
[2] Dapeng Wu, Y. T. Hou, Ya-Qin Zhang, Transporting Real- Time Video over the
Internet: Challenges and Approaches, Proc. of the IEEE , Vol. 88, No. 12 , pp.
1855 -1877, Dec. 2000
[3] Frank G. Lin, Transparent MPEG Video Filtering for Wireless Networks,
Masters thesis, University of Waterloo, Canada, 2000.
[4] S. Floyd, and V. Jacobson, Random Early Detection Gateways for Congestion
Avoidance, IEEE/ACM Trans. on Networking, V.1 N.4, pp. 397-413,Aug. 1993.
[5] G. Blakowski, R. Steinmetz, A Media Synchronization Survey: Reference Model,
Specification, and Case Studies, IEEE Journal on Selected Areas in
Communications, Vol. 14, No. 1, pp. 5 -35, Jan. 1996
[6] H.M. Radha, M. van der Schaar, Yingwei Chen, The MPEG-4 Fine- grained
Scalable Video Coding Method for Multimedia Streaming over IP, IEEE Trans.
on Multimedia, Vol. 3, No. 1 , pp. 53 -68, Mar. 2001
[7] Weiping Li, Overview of Fine Granularity Scalability in MPEG-4 Video
Standard, IEEE Trans. on Circuits and Systems for Video Technology, Vol. 11,
No. 3 , pp. 301 -317, Mar. 2001
[8] Weiping Li, Fan Ling, Xuemin Chen, Fine Granularity Scalability in MPEG-4
for Streaming Video, Proc. IEEE Int. Symp. on Circuits and Systems, Vol. 1, pp.
299 -302 , 2000
[9] Transparent End-to-end Packet Switched Streaming Service (PSS); Protocols
and codecs, 3GPP TS 26.234 V5.2.0, Sept. 2002
[10] Sumit Roy, Bo Shen, Vijay Sundaram, Application Level Hand-off Support for
Mobile Media Transcoding Sessions, Pro. of the 12th Int. Workshop on Network
and Operating Systems Support for Digital Audio and Video, May 2002
129
[11] Roger Karrer, Thomas Gross, Dynamic Handoff of Multimedia Streams, Proc.
of the 11th Int. Workshop on Network and Operating Systems Support for Digital
Audio and Video, Jan. 2001
[12] J. Kangasharju, F. Hartanto, M. Reisslein, K.W. Ross, Distributing Layered
Encoded Video through Caches, IEEE Trans. on Computers, Vol. 51, No. 6, pp.
622 -636, June 2002
[13] A.K. Katsaggelos, L.P. Kondi, F.W. Meier, J. Ostermann, G.M. Schuster,
MPEG-4 and Rate-distortion-based Shape-coding Techniques, Proc. of the
IEEE , Vol. 86, No. 6, pp. 1126 -1154, June 1998
[14] H. Schulzrinne, A. Rao, R. Lanphier, Real Time Streaming Protocol (RTSP) ,
IETF RFC 2326, Apr. 1998.
[15] H. Schulzrinne, S. Casner, R. Frederick, V. Jacobson, RTP: A Transport Protocol
for Real-Time Applications, IETF RFC 1889, Jan. 1996.
[16] Jiangchuan Liu, Huai-Rong Shao, Bo Li, Wenwu Zhu, Ya-Qin Zhang, A
Scalable Bit-stream Packetization and Adaptive Rate Control Framework for
Object-based Video Multicast, Proc. of IEEE Global Telecommunications Conf.,
Vol. 3, pp. 2020 -2025, 2001
[17] Feng Wu, Shipeng Li, Ya-Qin Zhang, A Framework for Efficient Progressive
Fine Granularity Scalable Video Coding, IEEE Trans. on Circuits and Systems
for Video Technology, Vol. 11, No. 3, pp. 332 -344, Mar. 2001
[18] S. Wenger, G.D. Knorr, J. Ott, F. Kossentini, Error Resilience Support in
H.263+, IEEE Trans. on Circuits and Systems for Video Technology, Vol. 8, No.
7, pp. 867 -877, Nov. 1998
[19] Dapeng Wu, Y.T. Hou, Ya-Qin Zhang, Scalable Video Transport over Wireless
IP Networks, Proc. of IEEE Int. Symp. on Indoor and Mobile Radio
Communications, Vol. 2, pp. 1185 -1191, 2000
[20] R. Aravind, M.R. Civanlar, A.R. Reibman, Packet Loss Resilience of MPEG-2
Scalable Video Coding Algorithms, IEEE Trans. on Circuits and Systems for
Video Technology, Vol. 6, No. 5, pp. 426 -435, Oct. 1996
130
[21] A. Ortego, K. Ramchandran, Rate-distortion Methods for Image and Video
Compression, IEEE Signal Processing Magazine, Vol. 15, No. 6, pp. 23 -50,
Nov. 1998
[22] J. Padhye, V. Firoiu, D.F. Towsley, J.F. Kurose, Modeling TCP Reno
Performance: A Simple Model and its Empirical Validation, IEEE Trans. on
Networking, Vol. 8, No. 2, pp. 133 -145, Apr. 2000
[23] Qian Zhang, Wenwu Zhu, Ya-Qin Zhang, Network-adaptive Scalable Video
Streaming over 3G Wireless Network, Proc. of IEEE Int. Conf. on Image
Processing, Vol. 2, pp. 579 -582, 2001
[24] S. Floyd, K.Fall, Promoting the Use of End-to-end Congestion Control in the
Internet, IEEE Trans. on Networking, Vol. 7, No. 4, pp. 458 -472, Aug. 1999
[25] W.E. Leland, M.S. Taqqu, W. Willinger, D.V. Wilson, On the Self-similar
Nature of Ethernet Traffic, IEEE Trans. on Networking, Vol. 2, No. 1, pp. 1 -15,
Feb. 1994
[26] John G. Apostolopoulos, Susie J. Wee, Unbalanced Multiple Description Video
Communication Using Path Diversity, Proc. of Int. Conf. on Image Processing,
Oct. 2001
[27] M. Gallant, F. Kossentini, Rate-distortion Optimized Layered Coding With
Unequal Error Protection for Robust Internet Video, IEEE Trans. on Circuits
and Systems for Video Technology, Vol. 11, No. 3, pp. 357 -372, Mar. 2001
[28] Yao Wang, L. Karam, G.P. Abousleman, T. Key, B. Razzouk, Error Resilient
Video Coding Techniques, IEEE Signal Processing Magazine, Vol. 17, No. 4, pp.
61 -82, July 2000
[29] I. Moccagatta, S. Soudagar, J. Liang, H. Chen, Error-resilient Coding in JPEG-
2000 and MPEG-4, IEEE Journal on Selected Areas in Communications, Vol. 18,
No. 6, pp. 899 -914, June 2000
[30] R. Talluri, Error-resilient Video Coding in the ISO MPEG-4 Standard, IEEE
Communications Magazine, Vol. 36, No. 6, pp. 112 -119, June 1998
[31] G. Cote, B. Erol, M. Gallant, F. Kossentini, H.263+: Video Coding at Low Bit
Rates, IEEE Trans. on Circuits and Systems for Video Technology, Vol. 8, No. 7,
pp. 849 -866, Nov. 1998
131
[32] A.S. Tosun, W.C. Feng, Efficient Multi- layer Coding and Encryption MPEG
Video Stream, Proc. of IEEE Int. Conf. on Multimedia and Expo, Vol. 1, pp. 119
-122, 2000
[33] S. Dogan, S. Eminsoy, A.H. Sadka, A.M. Kondoz, Personalised Multimedia
Services for Real-time Video over 3G Mobile Networks, Proc. of Third Int.
Conf. on 3G Mobile Communication Technologies, pp. 366 -370, 2002
[34] Yao Wang, Qin-Fan Zhu, Error Control and Concealment for Video
Communication: A Review, Proc. of the IEEE , Vol. 86, No. 5, pp. 974 -997,
May 1998
[35] M. B. Jill, Robert D. Gaglianello, Packet Loss Effects on MPEG Video Sent
Over the Public Internet, ACM Multimedia 98 - Electronic Proc., Sept. 1998
[36] Dapeng Wu, Y.T. Hou, Wenwu Zhu, Hung-Ju Lee, Tihao Chiang, Ya-Qin Zhang,
H.J. Chao, On End-to-end Architecture for Transporting MPEG-4 Video over the
Internet, IEEE Trans. on Circuits and Systems for Video Technology, Vol. 10,
No. 6, pp. 923 -941, Sept. 2000
[37] Chung-Ming Huang, Ming-Yuhe Jang, Handoff Architectures and Protocols for
Transmitting Compressed Multimedia Information in Mobile PCSs, IEEE Trans.
on PCSs, Consumer Electronics, Vol. 43, No. 3, pp. 784 -794, Aug. 1997
[38] Y.Wang, Jrn Ostermann, Ya-Qin Zhang, Video Processing and Communications,
Prentice Hall, 2002
[39] J. Heinanen, F. Baker, W. Weiss, J. Wroclawski, Assured Forwarding PHB
Group, IETF RFC 2597, June 1999.
[40] D. Black, Differentiated Services and Tunnels, IETF RFC 2983, Oct. 2000.
[41] B. Davie , A. Charny, J.C.R. Bennett, K. Benson, J.Y. Le Boudec, W. Courtney,
et al, An Expedited Forwarding PHB (Per-Hop Behavior) , IETF RFC 3246,
Mar. 2002.
[42] N. Brady, MPEG-4 Standardized Methods for the Compression of Arbitrarily
Shaped Video Objects, IEEE Trans. on Circuits and Systems for Video
Technology, Vol. 9, No. 8, pp. 1170 -1189, Dec. 1999
[43] J. Apostolopoulos, T. Wong, Wai-tian Tan, S. Wee, On Multiple Description
Streaming with Content Delivery Networks, Proc. of Twenty-First Annual Joint
132
Conf. of the IEEE Computer and Communications Societies, Vol. 3, pp. 1736 -
1745, 2002
[44] K. Nichols, S. Blake, F. Baker, D. Black, Definition of the Differentiated
Services Field (DS Field) in the IPv4 and IPv6 Headers, IETF RFC 2474, Dec.
1998.
[45] S. Blake, D. Black, M. Carlson, E. Davies, Z. Wang, W. Weiss, An Architecture
for Differentiated Services, IETF RFC 2475, Dec.1998.
[46] J. Apostolopoulos, Error-Resilient Video Compression Through the Use of
Multiple States, Proc. of Conf. on Image Processing, Sept. 2000
[47] J. Apostolopoulos, Reliable Video Communication over Lossy Packet Networks
Using Multiple State Encoding and Path Diversity, Proc. of Visual
Communications and Image Processing, pp. 392-409, Jan. 2001
[48] S. Nanda, K. Balachandran, S. Kumar, Adaptation Techniques in Wireless
Packet Data Services, IEEE Commun. Mag., vol. 38, pp.5464, Jan. 2000
[49] V.K. Goyal, Multiple Description Coding: Compression Meets the Network,
IEEE Signal Processing Magazine, Vol. 18, No. 5, pp. 74 -93, Sept. 2001
[50] IP in the RAN as a transport option in 3
rd
generation mobile systems, Mobile
Wireless Internet Forum MTR-006 v2.0.0, June 2001
[51] General Packet Radio Service (GPRS); GPRS Tunnelling Protocol (GTP) across
the Gn and Gp Interface, 3GPP TS 29.060 V5.4.0, Dec. 2002
[52] John W. Mark and Weihua Zhuang, Wireless Communications and Networking,
Prentice Hall, New Jersey, 2002
[53] Qian Zhang, Wenwu Zhu, Ya-Qin Zhang, Resource Allocation for Multimedia
Streaming over the Internet, IEEE Trans. on Multimedia, Vol. 3, No. 3 , pp. 339
-355, Sept. 2001
[54] Y. Bernet, S. Blake, D. Grossman, A. Smith, An Informal Management Model
for Diffserv Routers, IETF RFC 3290, May 2002
[55] J. Schiller, Mobile Communications, Addison-Wesley, 2000
[56] Combined GSM and Mobile IP Mobility Handling in UMTS IP CN, 3G TR
23.923 V.3.0.0, May 2000
133
[57] Handovers for real-time services from PS domain, 3GPP TR 25.936 V4.0.1,
Dec. 2001
[58] Handover procedures, ETSI TS 100 527 V7.0.0, Aug. 1999
[59] Handover procedures, 3GPP TS 23.009 V5.3.0, Dec. 2002
[60] Network architecture, 3GPP TS 23.002 V5.9.0, Dec. 2002
[61] Quality of Service (QoS) concept and architecture, 3GPP TS 23.107 V5.7.0,
Dec. 2002
[62] J. Wiljakka, Transition to IPv6 in GPRS and WCDMA Mobile Networks, IEEE
Communications Magazine, Vol. 40, No. 4, pp. 134 -140, Apr. 2002
[63] Wei-Ge Chen, Ming-Chieh Lee, -channel Compression in Video Coding,
Proc. of IEEE Int. Conf. on Image Processing, Vol. 1, pp. 500 -503, 1997
[64] J. De Vriendt, P. Laine, C. Lerouge, Xiaofeng Xu, Mobile Network Evolution: A
Revolution on the Move, IEEE Communications Magazine, Vol. 40, No. 4, pp.
104 -111, Apr. 2002
[65] David Leon, RTP Retransmission Framework, IETF Internet draft, Mar. 2002
[66] J. Widmer, R. Denda, M. Mauve, A Survey on TCP-friendly Congestion
Control, IEEE Network, Vol. 15, No. 3, pp. 28 -37, May-June 2001
[67] General Packet Radio Service (GPRS); Service description; Stage 2, 3GPP TS
23.060 V5.4.0, Dec. 2002
[68] S. Das, A. Misra, P. Agrawal, TeleMIP: Telecommunications- Enhanced Mobile
IP Architecture for Fast Intradomain Mobility, IEEE Personal Communications,
Vol. 7, No. 4, pp. 50 -58, Aug. 2000
[69] F.M. Chiussi, D.A. Khotimsky, S. Krishnan, Mobility Management in Third-
Generation All-IP Networks, IEEE Communications Magazine, Vol. 40, No. 9,
pp. 124 -135, Sept. 2002
[70] F.A. Chiussi, D.A. Khotimsky, S. Krishnan, A Network Architecture for MPLS-
based Micro-mobility, Proc. of Wireless Communications and Networking Conf.,
Vol. 2, pp. 549 -555, 2002
[71] T. Ahmed, A. Mehaoua, G. Buridant, Implementing MPEG-4 Video on Demand
over IP Differentiated Services, Proc. of IEEE Global Telecommunications Conf.,
Vol. 4, pp. 2489 -2493, 2001
134
[72] T. Ahmed, G. Buridant, A. Mehaoua, Encapsulation and Marking of MPEG-4
Video over IP Differentiated Services, Proc. of IEEE Symp. on Computers and
Communications, pp. 346 -352, 2001
[73] Jitae Shin, Jong Won Kim, C.C.J. Kuo, Quality-of-service Mapping Mechanism
for Packet Video in Differentiated Services Network, IEEE Trans. on
Multimedia, Vol. 3, No. 2, pp. 219 -231, June 2001
[74] UTRAN Overall Description, 3GPP TS 25.401 V5.5.0, Dec. 2002
[75] T. Yoshimura, T. Ohya, T. Kawahara, M. Etoh, Rate and Robustness Control
with RTP Monitoring Agent for Mobile Multimedia Streaming, Proc. of IEEE
Int. Conf. on Communications, Vol. 4, pp. 2513 -2517, 2002
[76] R.Bennett and H.Zhang, WF
2
Q: Worst-case Fair Weighted Fair Queuing, Proc.
of IEEE INFOCOMM '96, pp. 120-128, March 1996.
[77] OPNET Modeler 9.0 Documentation, OPNET Technologies, Inc., 2002.