Performance Analysis of Soft and Hard Handovers

based on UMTS QoS Traffic Classes

Muhammad Naveed Iqbal Ajmal Khan
Department of Software Engineering, Department of Computer Sysgtems Engineering,
University of Engineering & Technology, Mardan Campus, University of Engineering & Technoloyg, Peshawar, Pakistan
Pakistan engajmal@yahoo.com
mohammadnaveediqbal@yahoo.com

Abstract—The Universal Mobile Telecommunication There are many reasons in which the UMTS network
system (UMTS) is one of the emerging cellular phone differs from prior systems. One reason is that previous
technologies which are known as the 3G system. It support cellular systems were mainly circuit-switched, while
the high speed data transfer, speech, web browsing, email, UMTS is packet-switched. It also has higher bandwidth
video telephony, multimedia and the audio streaming. than previous systems [3].
These services are divided in to the classes depending upon
the QoS requirements. With the development of these There are four different QoS classes: Conversational
cellular networks, a major problem came up; it was the class, Streaming class, Interactive class and Background
handover from one cell to the other cell during an ongoing class [2]. The main distinguishing factor between these
session without dropping the connection with the base QoS classes is how delay sensitive the traffic is:
station. A lot of techniques were developed and used to Conversational class is meant for traffic which is very
cope with this major issue. The user’s movement is
dynamic process i.e. mobile user can change its way any
delay sensitive while Background class is the most delay
time with any speed, so there should be a mechanism that insensitive traffic class [4].
enables the network to be aware of this process. For this Conversational and Streaming classes are mainly
purpose different types of handovers techniques are used
intended to be used to carry real-time traffic flows. The
which include soft and hard handovers. The capability of a
cellular network to perform efficient handovers is essential main divider between them is how delay sensitive the
to offer attractive services. This research paper presents traffic is. Conversational real-time services, like video
the analysis of different handovers in the 3G UMTS telephony, are the most delay sensitive applications and
network which is the vital issue to the network to maintain these data streams should be carried in Conversational
the user’s connection during the ongoing session with the class [4].
user’s movement. The analysis is based on the UMTS QoS
traffic classes. For this purpose the soft and the hard Interactive and Background Classes are mainly
handovers techniques are analyzed in different scenarios meant to be used by traditional Internet applications like
which are implemented in the OPNET Modeler. www, Email, Telnet, FTP and News. Due to looser delay
requirements, compare to conversational and streaming
Keywords- UMTS, QoS, UE, RTP, WCDMA
classes, both provide better error rate by means of
I. INTRODUCTION channel coding and retransmission. The main difference
between Interactive and Background class is that
UMTS stands for the Universal Mobile
Interactive class is mainly used by interactive
Telecommunication System. The UMTS network is a
applications, e.g. interactive Email or interactive Web
wireless 3G (third generation) network that provides
browsing, while Background class is meant for
high bandwidth voice and data service to users of mobile
background traffic, e.g. background download of Emails
devices. The UMTS network is also called 3GSM
or background file downloading. Responsiveness of the
(Global System for Mobile communications) because it
interactive applications is ensured by separating
evolved from that system. The air interface for the
interactive and background applications. Traffic in the
UMTS network is based on WCDMA (Wideband Code
Interactive class has higher priority in scheduling than
Division Multiple Access) and includes the HSPA (High
Background class traffic, so background applications use
Speed Packet Access) specification [1].
transmission resources only when interactive
The 3G systems were created with the intention of applications do not need them. This is very important in
allowing users to have global mobility with services wireless environment where the bandwidth is low
including Internet, data, messaging, paging, and compared to fixed networks [4].
telephony. The idea was to provide consistent service to
This paper compares the two handover techniques
roaming mobile customers anywhere in the world. A
[6] in terms of four UMTS QoS classes. Two different
combination of terrestrial-based wireless services and
scenarios are implemented in OPNET Modeler [5]and
satellite transmissions were designed to provide this
simulation results are compared to decide which
availability [2].
handover scheme is best suited for these four UMTS all aspects was than duplicated and attributes were set
QoS classes. for both the scenarios. Now there were two scenarios
one for HARD HANDOVER and other for SOFT
The paper is organized as follows. In section II,
HANDOVER. Then the Simulation was completed for
UMTS architecture is discussed in brief. Detailed
different seed values and results were examined.
simulation results are explained in section III. Section IV
compares the performance of soft and hard handovers III. SIMULATION RESULTS
while section V concludes the research paper.
A number of different statistics are simulated for
II. UMTS ARCHITECTURE IN OPNET both soft and hard handover and resulting graphs of this
simulation are analysed on the basis of QoS classes that
The general architecture of UMTS is implemented in
which handover provides better services for these QoS
OPNET by using different nodes from the object palette
classes. The results concluded can be very much
as shown in Figure 1. IP cloud [7] was used between the
helpful for the network operators to provide better
servers and the core network [8] to show that a network
QoS to the end users. These analyzed results and
is involved. IP links were used to connect IP cloud to
conclusions will improve the quality of services in the
both the servers and the core network or GGSN node
next generations of mobile networks.
[9]. These nodes were placed in a fashion so that the
architecture can be clearly understood. All these nodes A. Analysis of Handover:
were connected using different connection links from the Three different statistics are observed to check
object Palette. whether the handover is being performed by the network
or not.
1) Active Cell Count: This statistic reports the
number of the cells in the Active Set of the
surrounding user equipment (UE), which varies during
soft handovers.

Figure 2: Active Set Cell Count

Figure 1: General UMTS Architecture
Initially UE is attached only to a single Node-B.
After the architecture was completed, the required
Therefore the statistic starts with an initial value of 1 for
attributes were set for each node. Applications were
the upper and lower part of the Figure 2. As during hard
defined in the application definition node and were
handover a cell replacement operation happens, the UE
assigned to the respective QoS of UMTS. The profiles
is attached to only a single Node-B at a time, it breaks
were created in the profile configuration node to be used
the connection from one cell and then establishes with
by the users.
the other, which does not actually change the count of
By using the Cell Creator utility, a grid of cells can the cells in the set, the statistic is recorded. The repeating
be drawn in the Project Editor. The Cell Creator utility statistic values i.e.1, in Figure 1 indicates Active Set cell
takes an existing map and superimposes on it a grid replacement events, which shows that hard handover is
structure with the parameters specified. It then creates a performed
new map that can be used in network topology. Figure 1
The lower part of the Figure 1 shows the soft
shows the cells created on the map.
handover. Whenever an addition or removal takes place
Trajectory for the user equipment was defined based to/from the Active Set, the new count of the cells is
on simulation time. A trajectory is a path specification recorded. The new count of the cells shows that a UE is
for a mobile site's motion during a simulation. In attached to more than one Node-B at a time as it
Modeller, a trajectory can be defined as either segment– happens in the soft handover procedure. As the
based or vector-based. The trajectory used is segment- handover in this scenario is performed between the two
based as shown in Figure 1. cells of two different Node-B’s as shown in Figure 1,
so before handover UE gets connected with both the
In order to analyze different results, global and node
cells and the cell count value becomes 2 and after the
statistics were selected. A single scenario completed in
good signal strength is received from the target cell the voice packet network delay is the time at which the
connection with previous cell is terminated which sender node gave the packet to real time transport
means UE is again connected to the one cell and the protocol (RTP) to the time the receiver got it from RTP.
Figure 5 shows that in hard handover there is more
cell count value becomes 1 again as shown in the delay in packets as compared to soft handover. This is
Figure2. because in hard handover loss of session occurs while in
soft handover the session is continued and not
2) Cells added to active set: This statistic reports terminated.
the Cell IDs of the cells that are added to the Active
Set of the surrounding UE throughout the simulation b) Traffic Sent (packets /sec): This statistic
initially and during handovers. Each cell ID reports Average number of packets per second
information is recorded at the simulation time when submitted to the transport layer by the voice
application. Figure 6 shows that there are same number
the addition takes place. This statistic is collected only of packets sent in both soft and hard handover because
for the UEs whose node type is mobile. As only two all the resources used and the environmental
cells are involved in the handover procedures with the conditions are same for both the scenarios.
ID’s 0 and 1 as shown in Figure 1, Figure3 shows the
ID’s of the cells added to the active set (either 0 or 1)
during the simulation time.
B. Analysis of Conversational Class (QoS 0):
The End to End Delay and voice traffic flow
statistics for conversational class are simulated using
both soft and hard

Figure 5: Packet end to end delay (sec)

Figure 3: cells added to active set

handover schemes. Figure 4 shows that there is less

end to end delay for QoS 0 in soft handover compared
to the hard handover. This delay is due to a fact that in
hard handover there is a loss of information because
Figure 6: Traffic Sent (packet/sec)
connection of UE is first terminated from the previous
cell and then it gets connected with the new or target c) Traffic Received (packets /sec): This statistic
cell. While in soft handover the user first gets report average number of packets per second forwarded
connected to the target cell and then terminates the to the voice application by the transport layer. Figure 7
previous connection. shows that traffic received in soft handover is greater
than in hard handover because there is no session loss
in soft handover and there is no loss of information.

Figure 4: End to End Delay

1) Voice Application: A number of different

Figure 7: Traffic Received (packet/sec)
statistics of voice application for QoS 0 are analyzed to
make some conclusions. 2) Conclusion for QoS 0: The QoS 0 also called
a) Packet End-To-End Delay (sec): The total conversational class has the characteristic of tolerating
 less delay i.e. it is very much delay sensitive. From the all the nodes in the network. Figure 10 showst that there
above discussion it is clear that the end to end delay of is less packet end to end delay in soft handover than
QoS 0 is more in hard handover as compared to the soft hard handover.because in hard handover loss of session
handover. Similarly the packet end to end delay is also occurs.This little delay is affordable for the QoS1.
more in hard handover. The traffic sent is almost the
same in both hard and soft handover but more traffic is
received in soft handover than in hard handover. So it
is concluded that the soft handover which has less
delay than hard handover, is suitable for the
conversational class.
C. Analysis Of Streaming Class (Qos 1):
The end to end delay and the application (streaming
video) traffic flows for streaming class are simulated
using both soft and hard handover schemes.
Figure 10: Packet end to end Delay
Figure 8 shows that there is less end to end delay for
QoS 1 in hard handover than soft handover. The c) Traffic Sent (packets /sec): It is the Average
streaming video application includes data and voice number of packets per second submitted to the transport
packets so there might be a delay due to the load, which layers by all video conferencing applications in the
when added to the delay produced by soft handover network. Figure 11 shows that there are same number
increases the overall delay. of packets sent in soft handover as in hard handover
but with a little delay.
1) Streaming Video: A number of different
statistics of streaming video application for QoS 1 are d) Traffic Received (packets /sec): Average
analyzed to make some conclusions. number of packets per second forwarded to all video
a) Packet Delay Variation: Variance among end conferencing applications by the transport layers in the
to end delays for video packets. End to end delay for a network. Figure 12 shows that more packets are
video received in hard handover but with a bit of delay in it .

Figure 8: End to End Delay Figure 11: Traffic sent (Packet / sec )

packet is measured from the time it is created to the

time it is received. Figure 9 shows that there is high
packet delay variation in soft handover than in hard
handover.

Figure 12: Traffic Received

2) Conclusion for QoS 1: The QoS 1 also called

streaming class has the characteristic of tolerating less
delay but it is less delay sensitive as compared to
Figure 9: Packet delay variation conversational class. From the above discusion it is
clear that the end to end delay of QoS 1 is more in soft
b) Packet End To End Delay (sec): This the time handover as compared to the hard handover. The packet
taken to send a video application packet to a destination end to end delay is more in hard handover. The traffic
node application layer. This statistic records data from sent is almost the same in both hard and soft with a litle
delay in case of hard handover but more traffic is
received in hard handover with a little bit delay which is
affordable in the streaming class. So it can be concluded
that the hard handover is suitable for the streaming class
because it receives more data with a little delay which is
affordable in streaming class.
D. Analysis Of Interactive Class (Qos 2):
The end to end delay and the application (Email and
Http) traffic flows for Interactive class are simulated
using both soft and hard handover schemes. Figure 15: Traffic sent (packet/sec)

Figure 13 shows that there is less end to end delay c) Traffic Received (packets/sec): Figure 16
for QoS 2 in hard handover than in soft handover shows that more traffic is received in hard handover
because in hard handover there is a connectivity with with less response time as compared to the soft
one Node-B and there is more delay in soft handover handover. As this QoS is less sensitive to delay so the
because it takes time for the selection of Node-B. delay of hard handover does not affect the reception and
1) Client Email: A number of different statistics of more data is received with hard handover.
Client Email application for QoS 2 are analyzed to d) Upload Response Time (sec): Time elapsed
make some conclusions. between sending emails to the email server and
a) Download Response Time (sec): It is the time receiving acknowledgements from the email server.
elapsed between sending request for emails and Figure 17 shows that upload response time is less for
receiving emails from email server in this node.Figure hard handover than in soft handover The upload
14 shows that the download response time is less in response time is more in soft due the reason that UE
hard handover than in soft handover The download sends data simultaneously to the two node-B’s and
response time is more in soft handover because UE while doing this it takes time in uploading.
receives data simultaneously from two node-B and
while making decision it takes time in downloading.

Figure 16: Traffic Received

Figure 13: End to End Delay

Figure 17: Upload response time

Figure 14: Client email download response time

2) Conclusion for QoS 2: The QoS 2 also called
b) Traffic Sent (packets/sec): Figure 15 shows interactive class has the characteristic that it is delay
that more traffic is sent in hard handover with less insensitive. The data integrity is important in this class
response time as compared to the soft handover. As this so UE using the QoS 2 applications cannot afford data
QoS is less sensitive to delay so the delay of hard loss in the transmission and reception. From the above
handover does not affect the transmission and more results it is clear that the end to end delay of QoS 2 is
data is sent with hard handover. more in soft handover as compared to the hard
handover. Similarly the download and upload responses
in the hard handover are also less than in soft handover.
From the traffic sent and received Figures it becomes Figure 19: Download response time
clear that while performing hard handover the UE b) Traffic Sent (packets/sec): Figure 20 shows
receives more traffic than in soft handover. In soft that more traffic is sent in soft handover with less
handover the user equipment while taking the decision response time as compared to the hard handover. As this
of connectivity with either of the two Node-B’s loses QoS is less sensitive to delay but data is important so it
packets and the response time also increases due to the cannot afford the delay of hard handover because data is
refreshing of the page. So it can be concluded that the lost in the loss of session.
hard hondover which has less delay and less uploading
and downloading response time for more data transfer is
suitable for the interactive class.
E. Analysis Of Background Class (Qos 3):
The end to end delay and the application (FTP and
Background Email) traffic flows for Background class
are simulated using both soft and hard handover
schemes.
Figure 18 shows greater delay for QoS 3 in hard
handover compared to the soft handover. This is Figure 20: Traffic Sent
because the connection is lost in hard handover which
results in loss of packets. This loss in information is not c) Traffic Received (packets/sec): Figure 21
affordable in this traffic class. So soft handover in shows that more traffic is received in soft handover with
which there is no loss of session gives less delay. less response time as compared to the hard handover.
As this QoS is insensitive to delay but data integrity is
1) Client File Transfer Protocol (FTP): A number important. So during soft handover UE receives more
of data with less response time.
different statistics of FTP for QoS 3 are analyzed to 2) Conclusions for QoS 3: The QoS 3 also called
make background class has the characteristic that it is delay
insensitive. The data integrity is important in this class
so

Figure 18: End to End delay

some conclusions. Figure 21: Traffic Received

a) Download Response Time (sec): Figure 19 UE using the QoS 3 cannot afford data loss in the
shows that the download response time is less in soft transmission and reception. From the above results it
handover than hard handover. The download response is clear that the end to end delay of QoS 3 is less in
time is more in hard handover for this QoS. As data is soft handover as compared to the hard handover.
more important and during hard handover there is a loss Similarly the download and upload response time in
of session and data might be lost. So due to this delay in the soft handover are also less as compared to hard
hard handover, remote server takes more time to handover. From the traffic sent and received Figures
response. it becomes clear that in soft handover the UE receives
more traffic than in hard handover. So from these
observations it can be concluded that the soft
hondover which has less delay and has less response
time for uploading and downloading the data than
hard handover, is suitable for the background class.
IV. PERFORMANCE ANALYSIS OF SOFT &
HARD HANDOVERS

Two different scenarios were implemented for the

two types of the UMTS handovers i.e. soft handover and
hard handover.
A. Performance of Soft Handover [2]. S. Baudet, C. Besset-Bathias, P. Frene, N. Giroux
In the scenario of Soft Handover, the user equipment “QoS Implementation in UMTS Networks”
communicates in parallel with different Node-B’s. So [3]. M. Franik, I. Sekula, “Simulation model of UMTS
the cell is added and removed in such a way that the UE packet switched data transfer over UTRAN”
and the UTRAN always remain connected. This ELMAR, 2005. 47th International Symposium
handover is also called as “make before break “. It is [4]. Jussi Laukkanen “UMTS Quality of Service
because the connection is made first to the other Node-B Concept and Architecture.
and releases the older connection after making the [5]. Opnet Technologies Inc.: Opnet Modeler , Version
connection to the target Node-B. 7.0B, Manual, 2000.
[6]. Bchini, T. Tabbane, N. Chaput, E. Tabbane, S.
This type of handover provides better services for the Beylot, A.-L “Hard handover vs soft handover in a
real time and the downloading services. As it is clear city” Computers and Communications, 2009. ISCC
from the analysis that the two QoS classes i.e. the 2009. IEEE Symposium
conversational class and the background class provide [7]. Yi-Bing Lin Ai-Chun Pang Yieh-Ran Haung
better services during soft handover as compared to hard Chlamtac, I. Nat. Chiao Tung Univ., Hsinchu, “An
handover. all-IP approach for UMTS third-generation mobile
F. Performance of Hard Handover networks”
[8]. Harmatos, J. Anal, “Planning of UMTS core
In the scenario of Hard Handover, the user
networks” The 13th IEEE International
equipment communicates with different Node-B’s but
Symposium on Personal, Indoor and Mobile Radio
not simultaneously. So the cell is added and removed in
Communications - PIMRC 2002
such a way that the UE and the UTRAN do not always
[9]. Mishra, A. Bradley, “Performance and architecture
remain connected. Thus the source connection is broken
of SGSN and GGSN of general packet radio
first and then the target connection is made, so this type
service (GPRS)” Global Telecommunications
of handover is also called as “break before make”.
Conference, 2001. GLOBECOM '01. IEEE
This type of handover provides better services for the
interactive online remote access services and the
streaming services. As it is clear from the analysis that
the two QoS classes i.e. the streaming class and the
interactive class provide better services during hard
handover as compared to soft handover.
V. CONCLUSIONS

This research paper presented the concept of

handovers in 3G universal mobile telecommunications
systems. Different scenarios were implemented in
OPNET and the results were discussed in detail. This
research work gives an impression of the main factors
which are being affected by the soft and the hard
handovers techniques in terms of the UMTS QoS
classes. To understand about the handover process
between the Node B and the user equipment, different
statistics were calculated. On these measurements it was
observed that how different types of the traffic are
affected. This was analyzed in terms of the upload and
download responses, end-to-end delay and the packet
delay variations, by different types of handover. The
results were concluded on the basis of QoS classes that
for which class which handover is suitable. These
conclusions can be implemented in the real networks for
the provision of better services. The conclusions will be
helpful in future to implement the handovers according
to the applications or the quality of services the user is
using.

REFERENCES
[1]. Stijn N.P Van Cauwenberge “Study of soft
handovers in UMTS”.