REPOSITORY > RESULTS

Abstract

Over the last decade we have been witnessing rapid development of telecommunication networks and
services, in particular in the filed of new wireless network technologies. On different markets users are able
to communicate using different wireless technologies. The first two generation of mobile networks (i.e.
NMT and GSM) were developed to provide voice services, while with the third generation (UMTS) the
ability for faster data transfer was enabled. Data can be transmitted also over other wireless technologies,
among which the most popular are Bluetooth and WLAN that are using fixed line to access to core
network. On fixed access with evolving xDSL and FTTx technologies, operators can offer higher and
higher data rates.
Operators, that are offering services on fixed lines, are migrating fast to all IP core networks. Mobile
operators are following them. As a result of such migration, networks will become more heterogeneous,
combining different access technologies connected to IP core network.
The availability of different services is increasing and users will have need to use them using any
network. Today there are techniques available that enable movement between different segments of a single
network. In the future, operators will need to enable movement also between different access technologies.
Techniques that support user movement within and between different networks are defined as mobility
management techniques.
Mobility management in homogeneous networks (i.e. networks that are using same technology), is
already available today. Heterogeneous networks of the future will support different technologies. Thus,
new terminals and services are to be developed in order to provide seamless movement between different
access networks.
From the point of view of protocol implementation, handover in heterogeneous network can be
performed at different OSI layers and this has been addressed in several studies. In our work we focused on
solutions that can be deployed easily in a real operator environment. SIP is used today in many operator
environments and has been selected as the primary signalling protocol in IMS (IP Multimedia Subsystem)
networks. As it runs on the application layer, SIP is independent of access technologies, which enables
roaming in the network of an operator that is not offering SIP services, as the network, in which the user
roams, is used just to access the application server in home network. Thus, we choose SIP protocol for
development of new procedures for mobility management in heterogeneous networks.
With migration to all-IP networks, all data streams are merged in a single core network, which can lead
to situation, where services have effect on each other. This is not problematic for applications that do not
have high requirements for quality of service (QoS), such is web browsing. On the other hand it can have
major impact on time critical application such as voice and video. Such applications require
implementation of mechanisms that provide sufficient level of QoS if operators want to ensure appropriate
level of user experience (QoE). In our work we selected IP telephony as the application, which is one of the
most time critical applications.
When implementing IP telephony, operators are obliged to provide some additional functionality that
can change the architecture of the IP telephony solution, due to regulatory requirements. Usually they add a
Session Border Controller (SBC) to their network. By adding SBC to the architecture, some practices can
be in conflict with SIP architectural principles. SIP-based SBCs typically handle both signalling and media.
When providing seamless handover in heterogeneous networks, for real-time applications in particular,
the ability to provide appropriate QoS in the target network is crucial. The most critical phase of handover
process is handover decision. If the decision for handover is made based only on SNR measurement (which
is usually the case in homogeneous networks), the user terminal would always handover to another network
when the predefined SNR threshold is exceeded. In the case of congestion in the target access network, the
service can be significantly degraded or could become unusable. Thus, we developed new procedure named
CAHP, which enables to perform handover efficiently, taking into account also the congestion status of the
target network. The procedure works with any two networks. In order to present the proposed procedure
clearly, we choose WLAN and HSPA network, between which the handovers were made. We assumed that
congestion can happen only in WLAN network.
The CAHP procedure consists of two algorithms: Pre-probe and Mid-probe algorithm. The first is used
for testing WLAN network prior handover and it starts when SNR, which stays as prerequisite for
handover, exceeds predefined threshold. As network capabilities can change also during its use, we defined
the second algorithm, named Mid-probe, which is used for congestion testing after handover to WLAN.
Among all parameters used in the proposed CAHP procedure the most important are Tpre and Tmid. Those
two parameters need to be set carefully as they affect the level of signalling overhead and speed of
detecting potential congestion. We defined two methods of setting those two parameters. In the first, named
CAHP-C, the parameters are set on constant value. This means that values of Tpre and Tmid parameters will
stay constant during simulation, equal for both parameters. In the second, named CAHP-A, the parameters
are set adaptively, which means that their value changes during the simulation according to current
utilisation of WLAN network.
In order to analyze and evaluate the proposed handover procedure, we developed a simulation model of
a telecommunication system, which was developed using the simulation tool OPNET Modeler. As we
decided for handover on the application layer, which is not supported by OPNET, we customized some
pre-defined process models that incorporate SIP procedures.
In order to evaluate the CAHP procedure, we prepared several scenarios, divided in three sets. The first
simulation set was used as a reference, where CAHP procedure was not used. With results of that set we
showed that in unreliable network significant degradation of service could happen if the decision is made
based on SNR ratio only. Such situation can happen in public networks that are usually used by more users,
which can lead to congestions in the network. Usage of those networks is mainly free of charge and thus
attractive also for IP telephony service. The problem arises as such networks do not support QoS
mechanisms for time critical applications. The results of the first set of simulation show, that QoS of IP
telephony service was degraded when the target networks became congested and measured delays were
above 1 s, which is totally unacceptable for real time applications. Thus, new mechanisms need to be
implemented to test congestion status of target networks. In the second simulation set we evaluated
CAHP-C procedure and show that the results were much better than in the reference scenario. However for
getting the best results a lot of signalization messages for congestion testing was exchanged, which could
led to an overload on SBCs. To lower the signalization overhead, we defined CAHP-A procedure. Its
evaluation was done in the third simulation set. Among scenarios in the second and third simulation set we
were looking for scenario with optimal parameters on four different ways. All showed that the CAHP-A
procedure was more appropriate.
In proposed procedure we used SIP protocol for sending messages, by which the congestion status of
network was tested. Such approach is completely independent from lower layers (i.e. transport, network,
connection and physical). Due to independence from protocols used in lower layers, operators, if they are
already offering SIP telephony, easily deploy our procedure to they network as SBCs and IP terminals
already support SIP protocol. By implementation of CAHP procedure only software upgrade is needed on
SBCs and user terminals. Besides this in our procedure it is also possible to include other parameters
(settings of the user) when making the decision for handover. If we would use lower layers for congestion
testing, this would lead into bigger changes of the application or even new application development.
Congestion testing with other protocols could be limited also with safety settings on SBC, where usually
only SIP and RTP traffic is allowed. Similar security limitations can be an issue also in the LAN network in
the user’s environment.
We can conclude, that by using solutions presented in this dissertation, the QoE can be significantly
improved when making handover in heterogeneous networks while the load on operators equipment is
minimized.