9
Case Study: Service
Quality Support in an
IP-based Cellular RAN
In this chapter, we shall study an IP-based Radio Access Net-
work (RAN) as an example of applying the technologies of the
preceding chapters. In the framework of preceding chapters, an
IP RAN can be considered to be a multi-service Internet access
domain supporting endpoint mobility. From the viewpoint of Diff-
Serv, mobility is handled on the link layer. The traffic engineering
framework of IETF is used to structure the example.
In what follows, we shall first study the motivation for using
IP-based transport in cellular radio access network. Next, IP RAN
transport architecture is explained. After that, an implementation
of service quality support in IP-based RAN is accounted for using
IETF’s traffic engineering framework.
The technologies listed here are generic to use of DiffServ-based
transport in IP based RAN. DiffServ transport will be used in
cellular backbones as well. Concentrating on IP-based RAN in this
chapter has the advantage of providing a relatively well-defined
case study of a multi-service access network. The same principles
can be applied in the backbone network, too, but QoS management
there is easier due to higher aggregation level of flows.
Implementing Service Quality in IP Networks Vilho R
¨
ais
¨
anen
2003 John Wiley & Sons, Ltd ISBN: 0-470-84793-X
276 SERVICE QUALITY SUPPORT IN AN IP-BASED CELLULAR RAN
Further optimizations for service quality support in IP-based
RANs are possible, but are not discussed here. The goal of the
present chapter is to provide a case study showing how traffic
engineering and other advanced IP technologies reviewed in pre-
vious chapters can be put together to implement service quality
in an IP-based RAN.
9.1 MOTIVATION FOR USING IP-BASED TRANSPORT
IN CELLULAR RAN
The primary aim of QoS mechanisms in any cellular RAN trans-
port is providing of quality support for services mapped to the
four 3GPP traffic classes described in Chapter 5. Additionally, the
QoS mechanisms used must also provide an adequate quality sup-
port control layer and management layer traffic. The easiest way of
supporting services with varying quality requirements is to reserve
capacity in the network according to the sum total of all traffic
classes during peak hour.
Traditionally, Radio Access Networks of GSM, GPRS, and 3G
networks have been built using Time Division Multiplexing (TDM)
links, based on an Erlang calculation for the peak hour traffic.
This has been an adequate solution for networks in which the end
user traffic consists almost solely of circuit-switched voice tele-
phony. Interactive voice has strict delay requirements, whereby
the network needs to be dimensioned to accommodate bursty data
traffic demand during peak hours. The dimensioning of telephony
networks for aggregated speech traffic, taking into account Pois-
son process-like arrival of connections in individual servers (base
stations) is an established discipline within traditional telephony
(cf., e.g., [McD00]). The problem with applying this approach to
a multi-service network domain is that it does not make best use
of the operator’s investment in transport capacity, especially con-
cerning the variety of services that need to be supported by third
generation mobile networks. For this reason, new alternatives have
been evaluated. In a study made by the Mobile Wireless Internet
Forum (MWIF), using IP-based transport in the RAN of different
third generation networks was found to be a viable option [IRT01].
Data-type applications have been adopted in the mobile net-
work, including browsing the Internet using Wireless Application
9.1 MOTIVATION FOR USING IP-BASED TRANSPORT IN CELLULAR RAN 277
Protocol (WAP). With my GPRS phone, I can initiate a connec-
tion and start browsing the news titles from Financial Times or
Helsingin Sanomat, latter being the largest daily newspaper in Fin-
land, within a few seconds. I have ample time to check for latest
news headlines during a 5-minute bus trip to work from home.
Othertraffictypes,whicharemadepossiblebythe3GPPQoS
framework for mobile networks, are streaming and data transfer,
such as uploading or downloading of digital photographs. Adding
pictures to text using Multimedia Messaging Standard (MMS) is
possible already with the “2.5G” GPRS networks. The set of sup-
ported end user services for 3G will consist of not only speech
and short message service (SMS), but to include also MMS, data,
and real-time content. As discussed in Chapter 3, the differences
in service quality requirements in a heterogeneous traffic mix can
be used in network dimensioning.
An IP-based transport network in the RAN still has to be dimen-
sioned to support delay-critical traffic during peak hours, in the
same way as with traditional telephony. For the less urgent traffic
types, however, no fixed capacity needs to be reserved, but instead
the multiplexing gain of packet switching can be leveraged to
obtain high utilization level in the network. The enabling technol-
ogy here is DiffServ, which makes possible sharing a single capac-
ity “pipe” by diverse class of applications with varying service
quality support requirements. Implementation of service quality
with IP and differentiated treatment is advantageous when com-
pared to the traditional concept of reserving capacity for different
services in the network. The exact benefits from using DiffServ-
based transport vary according to the precise combination of end
users’ services in the mobile network, as well as the access network
topology, but analyses indicate that savings can be up to tens of
percent as compared to traditional network dimensioning. Radio
access network being a major factor in cellular network CAPEX,
this fact translates to monetary savings for the operator.
In general, a Radio Access Network transport based on IP brings
with it several benefits:
• Less transport capacity is needed in the RAN.
• There are fewer protocol layers to be managed.
• Same type of transport hardware can be used in RAN as in
Internet backbone.
278 SERVICE QUALITY SUPPORT IN AN IP-BASED CELLULAR RAN
Leveraging the benefits of IP in a best possible manner requires
further technologies, which are not within the scope of this chap-
ter. Discussion about some of the related issues can be found
in [IRT01].
An IP-based RAN also has the benefit of allowing the natu-
ral incorporation of other access technologies apart from cellular
radio into a single multi-service, multi-access network. In such
a network, IP is the protocol tying different access technologies
together. This concept, sometimes called the All-IP network, is
illustrated in Figure 9.1. In addition to the WLAN access shown
in Figure 9.1, other possible access methods include ADSL/SDSL
and wireline Ethernet.
The usability of IP-based RAN is not limited to 3GPP networks,
but can be used in other mobility networks as well. Indeed,
the analysis of MWIF referred to previously covers both 3GPP
and 3GPP2 networks, the latter being CDMA2000 variant of the
International Mobile Telecommunications 2000 (IMT2000) third
generation mobile framework. In the MWIF study, also IntServ has
been included as a potential IP service quality support technology
in the RAN.
More generically, service quality support in mobile networks
has been discussed recently in [CZ02]. In that scheme, bandwidth
brokers in radio network access network border routers perform
admission control to SLAs in a DiffServ network, based on effec-
tive bandwidth approximation. Adaptive applications have been
found to make it possible to raise the utilization level of an access
network [CZ02], and indeed the adaptive bit rate AMR codec that
is used in the 3GPP systems is able to adjust to available bit rate.
Irrespective of the radio access technology used, having a proper
Operator
backbone
IP-based
RAN
WLAN
access
Public internet
Figure 9.1 A High-level view of an All-IP network
9.2 IP RAN TRANSPORT ARCHITECTURE 279
service quality support model is important for providing proper
service quality support in a multi-service environment [GC02].
9.2 IP RAN TRANSPORT ARCHITECTURE
We shall next take a brief look at 3GPP IP RAN transport architec-
ture for WCDMA radio access. The architecture will be extended
to cover evolution versions of GPRS radio technologies as well, but
these are not covered here. Before taking a look at architectures,
let us make a short excursion to the world of generic transport
architecture of a mobile network.
9.2.1 PLMN transport architecture
A generic cellular network, called Public Land Mobile Network
(PLMN) in GSM parlance, consists of three distinct transport parts:
• long-haul backbone;
• medium-capacity fibre transport;
• bandwidth-limited access links.
The high-level transport architecture domains are shown in
Figure 9.2.
The long-haul backbone is typically a general-purpose transport
network having traffic engineering capabilities adequate for high
traffic aggregation levels. The backbone connects the medium-
capacity access part of the mobile network to mobility servers
GPRS
roaming
exchange
High-capacity
backbone
Medium-capacity
fibre transport
BTS/
Node B
BTS/
Node B
Internet
domain
GGSN
SGSN
Mobility
servers
Cellular network
Other
PLMN
Bandwidth-limited
access links
Figure 9.2 Transport domains in a GPRS/3GPP network
280 SERVICE QUALITY SUPPORT IN AN IP-BASED CELLULAR RAN
(including SGSN and GGSN), to other cellular networks via GPRS
Roaming Exchange (GRE) network via SGSN, and to non-cellular
Internet domains via GGSN. As the name implies, GRE network
facilitates roaming between operators by providing service qual-
ity enhanced connectivity between mobility servers in different
PLMNs. We shall return later in this chapter to the topic of service
quality support towards Internet domains.
Medium-capacity fibre transport network delivers the traffic
between access links to the long-haul backbone. (Node B is
the canonical name for base station in WCDMA.) The medium-
capacity fibre transport can often be dimensioned with sufficient
capacity so that it will not be congested.
Bandwidth-limited access links connect BTSs and Node B’s to
fibre-based transport. The link layer technology used in this part
of the transport architecture depends on the environment. Typical
technologies include, leased fibre links and microwave links.
The bandwidth-limited part of RAN transport can be hierar-
chical in nature, resembling a tree with Node Bs as leaves, and
DiffServ-capable routers making up the branches of the tree trunk
(see Figure 9.3). For resiliency purposes, also loops may be used.
As the name implies, the bandwidth-limited links are typically of
limited capacity, and interface to the medium-capacity fibre links.
A single PLMN may encompass thousands or tens of thousands of
Node Bs, and an accordingly large number of narrow-bandwidth
links in RAN branches. The cost of entire RAN equipment makes
up the largest part of the cost of an entire PLMN, part of that
Node B
Node B
Node B
Node B
DiffServ
router
DiffServ
router
DiffServ
router
Fibre
transport
Figure 9.3 An illustration of RAN transport hierarchy
Note
: Frame combiner not shown
9.2 IP RAN TRANSPORT ARCHITECTURE 281
being attributable to the transport used. Even though the cost
of the actual transport may not be percentually the largest part
of the total cost, transport capacity can be limited by the avail-
ability of leased capacity or licences for the microwave transport.
Thus, transport capacity savings of tens of per cents are made
possible multiplexing gain using differentiation-based transport
inside RAN branches instead of per-aggregate capacity reserva-
tions translates significant transport cost savings for the mobile
network operator.
9.2.2 IP RAN transport architecture
IP-based RAN is an evolution of the 3GPP mobile network archi-
tecture, extending the scope of use of IP transport (Figure 9.4). In
GPRS and Release 99 3GPP architectures, the “core network” inter-
faces between GGSN and SGSN (including GRE interface) have
been based on IP. In Release 4, a 3GPP standardization successor
of Release 99, IP transport option up to Radio Network Controller
(RNC) was made possible. In the IP option of Release 4, the basic
R99 architecture is not modified, the only change being the trans-
port technology used beneath the GTP tunnel between the Node
B’s and RNC.
An IP RAN is a logical next step in this evolution, integrating the
IP-based transport more closely into the RAN architecture. It turns
out that for overall efficiency, it is beneficial to redesign the radio
network control layer, replacing a single RNC with distributed
radio resource control architecture. The explanation of the entire
IP RAN radio resource control architecture is not within the scope
of the book. For the purposes of this chapter it is sufficient to know
that from the viewpoint of mobile terminals, the 3GPP service
quality support control provides the same service quality support
User IP layer
Application
Radio tunneling
RAB
Link
Radio tunneling
RAB
Link
GTP
IP
Link
GTP
IP
Link
GTP
IP
Link
GTP
IP
Link
Link
User IP layer User IP layer
Terminal IP base transceiver
station & radio
network gateway
Serving GPRS support
node
Gateway GPRS support
node
Figure 9.4 IP RAN architecture protocol stacks for user layer traffic
Note
: The architecture has been simplified to make the role of bear-
ers clearer
282 SERVICE QUALITY SUPPORT IN AN IP-BASED CELLULAR RAN
as the preceding 3GPP network variants. An overview of the IP
RAN protocol stacks on a logical level is shown in Figure 9.4.
From the viewpoint of protocol stacks, the only difference in the
actual stacks is that the GTP tunnel uses IP-based bearer up to
basis station. Please note that in Figure 9.4, only user layer traffic
protocol stack is shown.
9.2.3 Handover traffic
Mobile networks based on Code Division Multiple Access (CDMA),
of which Wideband CDMA (WCDMA) used in 3GPP networks is
a subspecies, use the so-called soft handover for handling terminal
mobility. This means that a terminal may communicate with the
mobile network through more than one base station (called “Node
B” in CDMA) at a time. A mobile terminal does not have to detach
from the previous Node B before commencing communication with
the next one. This arrangement is necessary due to CDMA power
control. Due to this, traffic to a particular mobile participating to soft
handover is being transmitted via multiple Node Bs in both uplink
and downlink directions.
A practical implementation of soft handover requires the function
of a frame selector function in the network, which – for uplink direc-
tion – receives radio frames from different soft handover “legs”,
and combines the signals arrived via different routes into a single
PDU stream towards the core network (see Figure 9.5). The frame
selector is also known as a macrodiversity combiner, with the term
“microdiversity” being reserved for mobility handled by individ-
ual Node Bs. For downlink direction, the signal is split into different
paths and frame combining is performed in the terminal. In 3GPP
Terminal
Node B
Node B
Frame selector
Core
network
Figure 9.5 Principle of soft handover
Note
: Transport of radio frames is shown by dashed lines, and normal GTP
tunnelled traffic by a solid line
9.2 IP RAN TRANSPORT ARCHITECTURE 283
Release 99 networks, frame combining for uplink and frame split-
ting for downlink is done in RNCs.
The importance of the soft handover concept for the present
discussion is that in addition to 3GPP user layer and signalling
traffic, radio frames being part of soft handover traffic need to be
transported in the links of bandwidth-limited RAN. Soft handover
traffic has typically high forwarding priority, the frame-combining
algorithm having limitations for the allowed delay difference on
the different paths.
In addition to soft handover traffic, WCDMA networks in gen-
eral can also carry so-called drift traffic, consisting of non-processed
radio frames. In the case of 3GPP R99 network, drift traffic can
be transported between a drift RNC and a serving RNC.Inthis
case, the drift RNC forwards non-processed radio frames to the
serving RNC. Drift traffic, too, has high priority. In what follows,
generic references to handover traffic are made, covering both soft
handover and drift traffic.
9.2.4 Service mapping in IP RAN
The 3GPP QoS model, reviewed in Chapter 5, applies indepen-
dently of the transport technology used in the RAN. Thus, the
user layer QoS model used in IP RAN is the same as in Release
99 networks. What is different between IP RAN and R99 UTRAN
is the implementation of the transport part of the radio access
bearer. The implementation of service quality support in IP RAN
transport is described below on a generic level.
As in 3GPP R99, each application flow is associated with a PDP
context describing the negotiated QoS support for the flow. The
QoS attributes of a PDP context have been described in more detail
in Chapter 5, and for the present purposes we are interested in
the QoS attributes of the PDP context which may affect IP RAN
transport. These include:
• Traffic class (conversational, streaming, interactive, or back-
ground);
• Traffic Handling Priority (THP);
• Allocation/Retention Priority (ARP).
These QoS attributes are used in mapping the PDP context asso-
ciated with the application flow onto a Radio Bearer (RB) for the