SIMULATION ANALYSIS OF PACKET SCHEDULING ALGORITHM FOR
VOICE, WWW AND VIDEO STREAMING SERVICES IN UMTS
DOWNLINK FDD MODE
Marko Porjazoski, Borislav Popovski
Department of Telecommunications, Faculty of Electrical Engineering, University St.s Cyril and Methodius
Keywords: UMTS, WCDMA, Radio Resource Management (RRM), Packet Scheduling, DCH, DSCH
Abstract: UMTS provides a new and important featur
e allowing negotiation of the property of the radio bearer. We
have focused on data transmission in WCDMA systems using packet scheduling for DCH and DSCH, in
case when voice, video streaming and WWW services are engaged. Network exploiting the DCH and
DSCH as a data transport channels can provide higher throughput than a network without DSCH, if a good
combination of resource sharing between the DCH and DSCH is select.
1 INTRODUCTION
The third-generation system known as the Universal
Mobile Telecommunication System (UMTS)
introduces very variable data rates on the air
interface, as well as the independence of the radio
access infrastructure and the service platform.
UMTS allows user application to negotiate bearer
characteristics that are most appropriate for carrying
information. For users this makes available a wide
spectrum of services through the Wideband Code
Division Multiple Access (WCDMA). The variable
bit rate and variety of traffic on the air interface have
presented new possibilities for radio resource
utilization (Holma, Toskala, 2001)(Laiho et al.,
2001).
The expression Radio Resource Utilization
(R
RU) covers all functionality for handling the air
interface resources of a radio access network. These
functions together are responsible for supplying
optimum coverage, offering the maximum planned
capacity, guaranteeing the required quality of service
(QoS) and ensuring efficient use of physical and
transport resources (3GPP TS 25.211). The Radio
Resource Management (RRM) consists of Power
Control (PC). Handover control (HC), congestion
control (typically subdivided into admission control
(AC), load control (LC) and packet data scheduling)
and the resource manager (RM).
In this paper we are focused on simulation
anal
ysis of performances of packet scheduling
algorithm proposed in (Sallent et al., 2001). In the
simulations 7 omnidirectional
BS’s (Base Stations) are assumed with variable
num
ber of users uniformly distributed around the
scenario, using voice, video and WWW services.
The paper is organized as fallows: at the
begi
nning (section 2), we describe downlink
transport channels (Downlink Dedicated Channel-
DCH and Downlink Shared Channel-DSCH) used
for voice and data transmission in our simulations;
in third section some aspects of interference and
code management are described with the formulas
for interference and load factor estimation; section
four is description of packet scheduling algorithm to
be simulated; section five presents simulation
scenario, in sections six simulation results are
discussed and section seven concludes the paper.
2 TRANSPORT CHANNELS
UMTS have a reach set of dedicated channels and
shared channels. These channels can support
multimedia applications ranging from voice to best-
effort data.
Two transport channels caring a downlink
t
ransmission are (3GPP TS 25.211):
a) Dedicated Channel (DCH) – is available
exclusi
vely to a specific user. Dedicated Channel is
devoted to services with stringent transfer delay
requirements (voice and video). The transmission
rate of a DCH can be changed every 10 ms.
67
Porjazoski M. and Popovski B. (2004).
SIMULATION ANALYSIS OF PACKET SCHEDULING ALGORITHM FOR VOICE, WWW AND VIDEO STREAMING SERVICES IN UMTS DOWNLINK
FDD MODE.
In Proceedings of the First International Conference on E-Business and Telecommunication Networks, pages 67-73
DOI: 10.5220/0001396700670073
Copyright
c
SciTePress
b) Downlink Shared Channel (DSCH) – is
defined to support flexible multiplexing of bursty
data traffic (WWW) in WCDMA. DSCH is usually
used to support best-effort data services, as it cannot
provide guarantees for the service quality. Its
Transmission
capacity is divided up among several
users. The number of users multiplexed on DSCH
varies with time. Since the base station may transmit
to many users at one time, co-channel interference
exists.
Depending on the type of service to be
provided, transport channels should be managed and
allocated appropriately. In this paper we are focused
on conversational (voice), streaming (video) and
interactive (www) services, which quality
requirements may be expressed trough achieved bit
rate, percentage of lost packets, delay and the delay
jitter.
In order to differentiate quality levels for
streaming services, we assume two layered video
application that is characterized by two different
flows: a basic layer, with the minimum requirements
for a proper visualization, and an enhancement layer,
that contains additional information to improve the
quality of the received images. Data traffic from
basic layer will be transmitted through DCH
channel, along with conversational (voice) traffic,
while the enhancement layer will be transmitted only
if there is capacity in the DSCH channels. It is
assumed that the possible retransmissions of the
basic layer can be carried out in the DSCH channel
together with the enhancement layer, and having a
higher precedence than the latter.
In case of interactive (WWW) service users, data
is transmitted trough DSCH
.
3 INTERFERENCE AND CODE
MANAGEMENT
Radio Resource Management (RRM) covers all
functionalities for handling the air interface
resources of a radio access network. These functions
are responsible for supplying optimum coverage,
offering the maximum planned capacity,
guaranteeing the required QoS and ensuring efficient
use of physical and transport resources (Laiho et al.,
2001)(3GPP TS 23.107). Good resource allocation
schemes will aim to assign many as possible links
with adequate SIR to mobile users. Resource
assignment is restricted by the interference caused
by the BS and terminals when they start using
assigned resources. RRM well not assign resources
to a terminal if this assignment would cause
excessive interference to other users. Decision about
who should transmit and its transmission parameters
(transport format and power level) are the
responsibility of the packet scheduler.
For the n users transmitting simultaneously at a
given cell, the following inequality for i-th user must
be satisfied (Sallent et al., 2001):
r
N
E
dL
PP
IP
SF
dL
P
i
o
b
ip
TiT
othiN
i
ip
Ti
×++
×
)(
)(
ρ
(1)
(2)
=
=
n
i
TiT
PP
1
PT – base station transmitted power; PTi –
power devoted to i-th user; Ii-oth – intercell
interference observed by i-th user; Lp(di) – i-th user
path loss, r – channel coding rate; PN – background
noise; SFi – spreading factor, relating i-th user data
bit rate and chip rate; ρ - orthogonality between
codes used in downlink direction.
If we rearrange equation (1), we will obtain
condition that PTi have to satisfy:
ρ
ρ
+
×++
r
N
E
SF
dL
P
IP
dLP
i
o
b
i
ip
T
othiN
ipTi
)(
)(
(3)
Adding all n (users) inequalities, the total power
transmitted by the base station can be expressed as:
=
+
=
n
i
i
o
b
i
ip
DL
N
TT
r
N
E
SF
dL
P
PP
1
max
)(
)1(
ρ
η
(4)
where ηDL is load factor defined as (Holma,
Toskala, 2001):
=
<
+
×
+
=
n
i
i
o
b
i
T
ipothi
DL
N
E
SF
P
dLI
1
1
)(
ρ
ρ
η
. (5)
Another common restriction is the number of
available codewords that BSs can use. Beside
interference control, packet scheduler should
manage dynamical allocation of OVSF codes. The
system has SFmax orthogonal codes with maximum
ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS
68
spreading factor SFmax = 512. According to the
properties of these codes, there availability is
guaranteed when the Kraft’s inequality is satisfied
(Minn, Siu, 2000):
=
n
i
b
ib
SF
R
R
1
max
,
(6)
where Rb,i is number of bits in transport block (TB)
for i-th user and Rb is minimal number of bits in TB
(corresponding to spreading factor SFmax = 512).
4 ALGORITHM DESCRIPTION
The scheduling strategy used in our simulations is
presented in (Sallent et al., 2001). Algorithm is
divided in two parts:
Kraft’s inequality
n + 1 users
P
T
(n +1, t) < P
Tmax
η
(n +1, t) <
φ
granted
transmission
postponed
transmission
Y
Y
Y
TF > 0
TF = TF - 1
Y
N
N
N
N
4.1 Prioritization
All users intended to transmit information must be
classified according to a certain prioritization
criteria. Firs scheduler will order requests depending
on service class they belong to, from highest to
lowest priority level. Conversational service class
(voice) has highest priority. Second, in this
prioritization scale, is streaming video service.
WWW (interactive service) has lowest priority.
Second prioritization rule is based on number of
basic layer TB in user’s buffer waiting for
retransmission (This rule can be applied only for
DSCH allocation, because there is no TB to be
retransmitted in DCH). When two or more users
have the same number of TB to be (re)transmitted, a
third prioritization level based on the Service Credit
(SCr) is considered.
The SCr is associated with an active link or user
and it computes the difference between the bit rate
requested by the user and the bit rate that system
provides to him. So it calculates the amount of
service that system owes to the user. The SCr value
of each active connection must be updated every
TTI (Transport Time Interval), following the
expression:
)1()1()( += kNumTx
TB
R
kSCrkSCr
G
(7)
where SCr(k) is the Service Credit for TTI=k,
SCr(k-1) is the Service Credit in the previous TTI,
RG is the guaranteed bit rate measured in bits/TTI,
TB is the number of bits in Transport Block for the
considered RAB (radio Access Bearer) and
NumTx(k-1) is the number of successfully
transmitted Transport Blocks in the previous TTI.
4.2 Resource Allocation and
Availability Check
Once requests are ordered, the next step will be to
decide whether or not they are accepted for
transmission and which is the accepted TF. The
limitations dealing with interference and code
availability are taken into account in this phase. For
this purpose we have to estimate the expected load
factor and transmitted power level once all the
requests are accepted. So, the expected load factor in
system with n active transmissions in frame t can be
estimated using equation (8).
Figure 1 : Resource allocation process
=
+
×
+
=
n
i
i
o
b
i
T
ipothi
N
E
SF
P
dLtI
tn
1
)()1(
),(
~
ρ
ρ
η
. (8)
The expected power is given by:
SIMULATION ANALYSIS OF PACKET SCHEDULING ALGORITHM FOR VOICE, WWW AND VIDEO
STREAMING SERVICES
69
=
+
=
n
i
i
o
b
i
ip
N
T
r
N
E
SF
dL
tn
P
tnP
1
)(
)),(
~
1(
),(
~
ρ
η
(9)
Here we have to mention that some differences
between real and estimated load factor value may
occur, as a consequence of inaccuracies in the
measurement of the other-to-own-cell interference
and path loss.
Algorithm execution follows the flow shown on
Figure 1. Assuming n already granted transmissions
and initially selected TF for n+1 request, the Kraft’s
inequality is evaluated, the expected load factor is
compared with a threshold φ and expected
transmission power should be below the maximum
transmitted power. If all tree conditions are satisfied,
transmission is granted for this request during one
TTI, otherwise, the transport format is reduced by
one (i.e. transmission bit rate is reduced). If this is
not possible, the request should wait for the next
frame.
5 SIMULATION SCENARIO
The system model includes 7 omnidirectional base
stations. Distance between two neighboring base
stations is 1 km. Maximum transmitted power by the
base station is 43 dBm. Mobile users are uniformly
distributed in the scenario moving with speed of 50
km/h. At each position update (every TTI) we
assume that mobile user will change his direction to
left or to right in 45° with probability 0.1 for each
side, and probability 0.8 that he will stay on previous
course. Path loss model used in this simulation is
adopted from (3GPP TS 25 942) and path loss
calculations are made according to (10).
L = 128.1 + 37.6 Log10(R) (10)
R is a distance from BS to ME. Path loss
calculated buy (10) shall in no circumstances be less
than free space path loss – FSPL = 20log(4πR/λ) . If
during calculations L become smaller than FSPL,
then FSPL should be considered instead L as path
loss.
After L (FSPL) is calculated, log-normally
distributed shadowing (LogF) with standard
deviation of 10 dB should be added, so that the
resulting path loss is the following:
Pathloss = L + LogF (11)
Number of voice, video and www users is taken
to be on of the parameters which will be changing
during simulations.
Traffic generation model proposed in (Perez-
Romero, 2002)(3FPP TR 101 112-UMTS 30.03) is
used, including following parameters for:
- Conversational Service (Voice traffic): On-Off
model with 0.3 activity factor. In the active period
voice users generates 160 bits in 10 ms (16 kbps).
- Interactive Service (WWW traffic): Session
arrival process – Poisson process; Number of packet
call requests per session – geometrically distributed
with mean 5 calls per session; Reading time between
packet calls – geometrically distributed with a mean
33 [s]; Number of packets within a packet call –
geometrically distributed with a mean value 25; Inter
arrival time between packets – geometrically
distributed with a mean 0.0625 [s] (for 64 kbps);
Packet size – Pareto Distribution (with cut-off), max
packet size 66666 bytes with parameters α=1.1,
k=81.5.
- Video Streaming Basic and Enhancement
Service: CBR model; bit rate 32 kbps (each 40 ms
packet with 1280 bits is generated);
In our simulations we have adopted: 40 ms for
TTI (Transport Time interval), 320 bits Transport
Block Size for WWW and Video streaming service
and 160 bits Transport Block Size for conversational
service. DCH used for transmission of voice and
video streaming basic layer will have two transport
formats (TFs): TF0 – no transmission and TF1
transmission of 4 Transport Blocks (TBs) in TTI.
Transport formats for DSCH, used for transport of
WWW and video streaming enhancement layer, are
listed in Table 1.
Table 1: Transport Formats for Downlink Shared Channel
(DSCH)
TB sizes, bits 320 bits (payload) + 16 bits
(MAC/RLC header)
TF0, bits
0×320
TF1, bits
1×320 (8 Kb/s)
TF2, bits
2×320 (16 Kb/s)
TF3, bits
4×320 (32 Kb/s)
TF4, bits
8×320 (64 Kb/s)
TFS
TF5, bits
16×320 (128 Kb/s)
TTI, ms 40
Duration of this simulation corresponds to 3 min
in real time. Life time of voice and video services
packets was set to 1s and for www service to 10s. If
the packet remains in the buffer longer then his life
time it will be discarded.
ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS
70
6 RESULTS
Results from the simulation of packet scheduling
algorithm described in this paper are shown in the
figures below. We have considered average user bit
rate, packet loss, delay and delay jitter as measures
that will help as to estimate system behavior and
performances. Figures provide a clear view of
system behavior in regard of number of users in the
scenario and the load factor threshold. Load factor
and power estimations, as it's shown in equation (8)
and (9), are based on interference measured in
previous TTI.
As a result of prioritization of different service
classes, it is obvious that best performances system
will provide to voice users (as service class with
highest priority), then to video users and at the
finally to www users
.
As you can see from the results, average bit rate,
packet loss and delay for the voice users are nearly
constant irrespectively to number of video and www
users and load factor (unless a small degradation of
performances for load factor values around 1).
On the other hand, system performances
experienced by video streaming service users
(second level in prioritization scale), appears to be
more depend on number of voice users and load
factor. It can be a little confusing, the packet delay
decrease for load factor in the order of 1. The
explanation for this behavior is that number of lost
packets for video users tremendously increases, and
in calculation of average packet delay we don take
this packets into account. Real system behavior for
the video streaming service can be noticed from
average bit rate, packet loss and packet delay jitter.
And finally, www service is treated as best effort
service class. When there is available resources left
by voice and video users, it will be allocated to www
users. As a result of smaller granularity of transport
formats, www user traffic is more adaptive to
severe system conditions.
7 CONCLUSIONS
In this paper, the performance of packet scheduling
algorithm, used the WCDMA network area with
seven cells, exploiting the DCH and DSCH was
studied by simulations. It was shown that as the
traffic load is increasing, the network resource
utilization increases until it reach the maximum. If
load is further increased, the resource utilization
saturates and starts decreasing again.
For smaller load factor threshold values (0.8-
0.9), packet scheduler does not provide best
performances. Load factor is the limiting factor that
do not allow new transmissions to be granted. For
increased load threshold values (values from 0.9 to
0.95) system provides best performances: maximum
average user bit rate, minimal packet loss, delay and
jitter.
For higher numbers of users and load factor
values around 1, difference between estimated and
real value of load factor is growing, system is
becoming unstable, base station transmitted power
reach its maximum value faster and becomes main
obstacle in provisioning better performances.
The results suggest that there is an optimal set of
parameters (load factor values form 0.9 to 0.95 and
number of users around 150, 75 voice, 50 video and
25 www users), by which the network resources
utilization reaches the maximum.
3400
3500
3600
3700
3800
3900
4000
4100
4200
0.8 0.85 0.9 0.95 1
Load Factor
Voice users bit rate (bps)
75 voice users
100 voice users
125 voice users
150 voice users
Figure 2: Average bit rate for voice service users in
scenario with constant number of video streaming users
(50) and constant number of www users (25)
0
5000
10000
15000
20000
25000
0.8 0.85 0.9 0.95 1
Load Factor
video users bit rate (bps)
75 voice users
100 voice users
125 voice users
150 voice users
Figure 3: Average bit rate for 50 video streaming service
users (basic and enhancement layer) in scenario with
variable number of voice users and fixed number of www
users (25)
SIMULATION ANALYSIS OF PACKET SCHEDULING ALGORITHM FOR VOICE, WWW AND VIDEO
STREAMING SERVICES
71
0
200
400
600
800
1000
1200
0.8 0.85 0.9 0.95 1
Load Factor
www users bit rate (bps)
75 voice users
100 voice users
125 voice users
150 voice users
Figure 4: Average bit rate for 25 www service users in
scenario with variable number of voice users and fixed
number of video streaming service users (50)
3500
3600
3700
3800
3900
4000
4100
0.8 0.85 0.9 0.95 1
Load Factor
Voice users bit rate (bps)
25 video users
50 video users
75 video users
Figure 5: Average bit rate for 75 voice service users in
scenario with variable number of video streaming users
and fixed number of www service users (25)
0
5
10
15
20
25
30
0.8 0.85 0.9 0.95 1
Load Factor
Voice users Packet loss (%)
75 voice users
100 voice users
125 voice users
150 voice users
Figure 6: Packet loss for voice users in scenario with fixed
number of video streaming users (50) and number of
www service users (25)
60
65
70
75
80
85
90
95
100
0.8 0.85 0.9 0.95 1
Load factor
Video users packet loss
75 voice users
100 voice users
125 voice users
150 voice users
Figure 7: Packet loss for voice users in scenario with fixed
number of video streaming users (50) and number of
www service users (25)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.8 0.85 0.9 0.95 1
Load Factor
Average delay (s)
75 voice users
50 video users
25 www users
Figure 8: Average packet delay in scenario with 75 voice
service users, 50 video streaming users and 25 www
service users
0.2
0.21
0.22
0.23
0.24
0.25
0.8 0.85 0.9 0.95 1
Load Factor
video users delaj jitter s)
25 video users
50 video users
75 video users
Figure 9: Packet delay jitter for video streaming users in
scenario with variable number of voice service users, 50
video streaming users and 25 www service users
ICETE 2004 - WIRELESS COMMUNICATION SYSTEMS AND NETWORKS
72
REFERENCES
Holma H., Toskala A., 2001. WCDMA for UMTS, John
Wiley & Sons.
Laiho J., Wacher A., Novosad T., 2001. Radio Network
Planning and Optimisation for UMTS, John Wiley &
Sons.
3GPP TS 23.107. Quality of Service (QoS) concept and
architecture
3GPP TS 25.211. Physical channels and mapping of
transport channels onto physical channels (FDD)
Sallent O., Pérez-Romero J., Casadevall F., Agustí R.,
2001. An Emulator Framework for a New Radio
Resource Management for QoS guaranteed Services in
W-CDMA Systems, In: IEEE Journal on Selected
Areas in Communications, Vol.19, No. 10, October
2001, pp. 1893-1904.
Minn T., Siu K.Y., 2000. Dynamic Assignment of
Orthogonal Variable-Spreading-Factor Codes in W-
CDMA, In: IEEE Journal on Selected Area in
Communications, August 2000, pp. 1429-1440.
Perez-Romero J., Sallent O., Agusti R., 2002. Downlink
Packet Scahouling for Two-Layered Streaming Video
Service in UMTS, In: IST Mobile & Wireless
Telecommunication Summit 2002, Thessaloniki –
Greece, June 2002, pp. 212-216.
3FPP TR 101 112-UMTS 30.03. Selection procedures for
the choice of radio transmission technologies of the
UMTS
3GPP TS 25 942 . RF system scenarios
SIMULATION ANALYSIS OF PACKET SCHEDULING ALGORITHM FOR VOICE, WWW AND VIDEO
STREAMING SERVICES
73