Evolving Wireless Systems: Choosing a Migration Path
telecoms-mag.com
Today's wireless systems are inefficient at handling small, frequent data calls
and bursty IP traffic, and herein lies the main bottleneck for implementing
high-speed data services. With an increasing number of data users, more time
slots will be required with existing circuit-switched connections, yet the
network and radio capacity required to support such large amounts of bursty
traffic would make it uneconomic or impossible (because of limitations on the
number of physical radio channels available) to implement. Mobile operators
will soon have to decide which technology path to use in order to migrate their
second generation systems to the third generation networks of the future.
David Tade
The development of mobile access to the Internet and other advanced
multimedia services, is pushed greatly by the convergence of wireless
voice and data. Present-day mobile access to the Internet is slow;
although the GSM data rate at 9.6 kbps (36 kbps with compression)
single time slot is suitable for messaging and short file transfer, it is not
very convenient for Web browsing. However, data rates are
expected to increase significantly in the future.
The Early Options
High-speed circuit-switched data (HSCSD) is the next step in
developing mobile access to the Internet, and is now available.
HSCSD allows multiple time slots (up to four slots) to be used for the
data connection. Thus, multiples of 14.4 kbps and 9.6 kbps can be
offered to deliver data rates of up to 57.6 kbps.
HSCSD service is optimal for file transfer and applications that
require constant high bit rate and constant transmission delay. It can
be used identically for all applications of 9.6 kbps data (email, remote
LAN access and the Web) with bit rate corresponding to landline
modems. HSCSD will not be used heavily for the Internet because it
is not very cost-effective, although early adopters of HSCSD will
have the advantage of building a competitive advantage by satisfying
the high-end subscriber segment. HSCSD is currently in use by
Finland's GSM-based Sonera, on a system supplied by Nokia.
General packet radio service (GPRS) is expected to be introduced
this year. It will deliver variable data rates from 9 kbps up to 171.2
kbps, although the practical speed limit is likely to be 115 kbps.
GPRS will cover all the major functions, including point-to-point
transfer of user data, Internet and X.25 interworking, filtering
functionality for security reasons, volume-based charging tools, and
roaming between public land mobile networks. By 2000, GPRS is
expected to have added point-to-multipoint transfer, supplementary
services, and additional interworking functionality (for example, ISDN
and modem interworking). Many in the industry perceive GPRS to be
the smooth evolution path to packet services and universal mobile
telecommunications system (UMTS) services. GPRS has the
advantage over HSCSD because of its capability to connect users at
all times while in a call.
Enhanced data for GSM evolution (EDGE) is an enhanced
modulation for GSM and is currently undergoing study. It will offer
higher bit rates than GPRS per timeslot, at 48.8 kbps. The maximum
data rate that will be available in a TDMA frame will be 384 kbps.
EDGE will be particularly good for network operators that do not
have UMTS licences but wish to offer services similar to UMTS at
less cost.
Dataquest research suggests that operators and manufacturers rate
the importance and opportunities for GPRS significantly higher than
HSCSD, but most will implement both before end of the decade or
early in year 2000. Tariffing is the key driver as GPRS bills by volume
of data rather than by connection time and its bursty nature is
attractive for mobile applications -- for example, electronic commerce
-- rather than dedicated packet networks.
Evolving GSM Network Infrastructure
Last year wideband CDMA (W-CDMA) was chosen by Europe as
the air interface for its proposed third generation system. W-CDMA
was selected for the following reasons: changing bandwidth
requirements of 3G systems, access to advanced multimedia services
during a call and its inherent spectrum efficiency. Third generation
network operators expect to offer on-demand services any time and
anywhere. The expectation is that high data throughput and coverage
at all levels should be provided. The following data rates are
envisaged: no more than 2 Mbps to achieve the anticipated services in
an indoor environment; an average of 144 kbps to meet most of the
perceived needs of highly mobile users in wide area networks and 9.6
kbps for satellite coverage.
W-CDMA will coexist with GSM networks. This has a number of
distinct commercial advantages, which include protecting the
investment of existing GSM networks and hence the customer base,
and bolstering existing relationships with vendors for continuous
product development. Figure 4 illustrates a GSM-UMTS network.
The UMTS network will provide all of the third generation capabilities
with the help of a dual-mode terminal. The network will fully support
handover from one system to another.
The path to providing high-speed wireless data is not clear cut. In
deciding which path is best, an operator first has to build a business
case for its investment. Presently, the market for wireless data is a
difficult one to predict. As a result, the cost required in upgrading
existing systems for high-speed data is uncertain. Figure 5 shows the
paths an operator may decide to implement for wireless data.
The Nokia HSCSD solution requires only a software upgrade to the
existing GSM network. HSCSD services can be introduced without
major investment in the existing network infrastructure -- no extensive
network element modification is required. The result is a fast
return-on-investment for the operator and faster data connections for
the end user. However, for operators that eventually want to offer
wideband service this solution may become a waste of investment.
GPRS is the key to packet-based services, and is the first step in
evolving GSM networks to 3G capabilities. New nodes will be
introduced to integrate GPRS in a GSM network: GPRS support
node (SGSN) and gateway GPRS support node (GGSN). The
SGSN will handle packet routing and call authentication including
mobility management. Traffic will be routed from the SGSN to the
base station controller, and to the mobile station via the base station.
The GGSN will provide access to Internet service providers (ISPs)
as well as the allocation of Internet protocol (IP) addresses.
Moving from GSM to HSCSD/GPRS
Data over switched-circuit GSM networks has been available for
several years, providing data rates of 9.6 kbps per dialled connection.
ETSI GSM phase 2+ considers data-throughput enhancements by
means of HSCSD. The first step is the current 9.6 kbps data rate
upgraded to 14.4 kbps, dedicated to individual dial-in users. This puts
GSM circuit-switched access on a par with GPRS in terms of data
throughput per base station coverage area. HSCSD also allows for
channel aggregation, providing access at speeds of up to 57.6 kbps
per user.
GPRS available bandwidth per channel depends upon which coding
scheme is used: CS1 provides connectivity under ‘all conditions' and
delivers a user throughput of up to 9.05 kbps, while CS4 requires
excellent radio signal quality (carrier-to-interference ratio of 27 dB),
and delivers a user throughput of up to 21.4 kbps. Today's GSM
networks are generally designed for a C/I of 9 dB to 11 dB, which
makes coding scheme CS2 attractive with its throughput of up to 13.4
kbps. Channels within a cell can be aggregated for higher data
throughput rates. The bandwidth per channel is shared among all
concurrent users of the service within that particular coverage cell. A
maximum utilisation rate of 75 per cent has been suggested in a paper
from Siemens mobile networks, and UK analysts Dataquest,
considers this a not unreasonable maximum utilisation rate for a
shared packet network.
A fully configured GPRS network could provide up to 112 kbps
shared bandwidth per cell, and up to eight cells per base transceiver
station (BTS). In practice, available bandwidth is likely to be much
lower, typically using coding scheme CS2 and operating four cells per
BTS. This would result in a single user gaining access of up to 70
kbps throughput, providing there are no other concurrent users
allocated to the operating cell. A typical BTS would have a total
packet data bandwidth of up to 281 kbps. Today's remote LAN
access and Web surfing applications are characterised by session
times longer than an average voice call. Although one part of a session
may gain the advantage of maximum data throughput, another part will
be delayed while waiting for a pause in the packet flow of another
user in the same cell. Hence, for such applications it is perhaps more
appropriate to consider average throughput per user. In a typical
GSM network, a BTS covers an average area of 2KM2. A fully
configured practical GPRS network would support up to eight
concurrent users per BTS coverage area at V.34 modem speeds, or
up to four concurrent users at an ISDN B channel speed of 64 kbps.
The number of subscribers per BTS will depend upon the acceptable
average data throughput rates and delays. Some studies on shared
media data communications suggest that a ratio of one user to 10 or
20 subscribers is acceptable for busy periods of the day, given a
good level of service. A GPRS network could support a maximum of
160 subscribers per BTS coverage area with an average level of
service similar to V.34 modems. This compares with 700 voice
subscribers on a GSM voice network.
The deployment of a new network dedicated to GPRS data services
should therefore take account of the premium over GSM voice tariffs
that such services would need to charge.
A more probable rollout of GPRS would be from existing GSM
networks where some of the GSM dial-access capacity will be
migrated to GPRS packet data access. This could be done by
allocating one cell per BTS to GPRS, or by allocating one or two
channels to GPRS in all cells. The former approach gives a higher
maximum data throughput, the latter a fairer share of the bandwidth to
users. With the network configured in a typical 4/12 cell reuse
scheme, the number of subscribers for a remote LAN access service
over GPRS would fall to about 40 per BTS, with no more than two
concurrent users for reasonable data throughputs.
From GSM to EDGE via GPRS
EDGE uses a different and more efficient modulation scheme: 16
quadrature amplitude modulation (QAM), rather than the gaussian
modulation shift keying (GMSK) scheme used over the radio
interface by GSM and GPRS. The 16 QAM system opens up more
bandwidth per radio carrier or cell. EDGE claims to provide data
rates of up to 384 kbps per cell, although this assumes that all eight
radio channels (time slots) are used and that one of the time slots is
not reserved for signalling. As with GPRS, the 384 kbps bandwidth
would be shared by all concurrent users operating within the same
cell.
The catch is that EDGE requires higher radio signal quality than that
found in an average GSM network before higher data throughput
speeds can be reached. This means more BTSs and infrastructure
build-out for established GSM operators that wish to migrate to
EDGE. The compatibility with GSM allows EDGE to be rolled out
gradually, with users stepping down to GPRS speeds in areas where
EDGE signal strength is insufficient.
Moving from GSM to UMTS Networks
Without a doubt, UMTS networks will be based on GSM as a core
network as already illustrated in Figure 2. However, the likely
characteristics of UMTS networks are still under standardisation.
UMTS terrestrial radio access network (UTRAN) will be connected
to the GSM-UMTS core network using a multi-vendor interface
referred to as Iu with W-CDMA as the air interface. The
GSM-UMTS network will consist of three main parts: GSM-UMTS
core network, URAN and the GSM base station system. There will
be two parts to the GSM-UMTS network: a circuit-switched part
and packet-switched part based on the GGSN (Figure 6).
In addition to the wideband services provided by UMTS, the use of
intelligent network (IN), customised application for mobile enhanced
logic (CAMEL), wireless access protocol (WAP) and telephony
value-added-services (TeleVAS), will create new services beyond
GSM phase 2+.
Cost of Implementation
The cost of implementing any of these wireless data solutions is a case
of ‘horses for courses'. Operators have varying needs, and their
deployment strategies will also vary. However, in providing
high-speed data service such as UMTS, there will be inherent costs
incurred by the operator depending on the chosen path (Table 1).
In the case of migrating from GSM to UMTS via GPRS or EDGE, it
is assumed that the initial transport protocol required is already being
built out and that operators have chosen IP switches rather than
asynchronous transfer mode (ATM).
The Transport Network
High data traffic -- in particular IP traffic -- is expected with UMTS
services. To accommodate the high traffic demand, a number of
factors must be considered in choosing a transport protocol, such as:
bandwidth efficiency; quality of service; speech delay-sensitivity;
standardisation stability and permitted maximum number of concurrent
users. ATM and IP switches are the only two protocols available on
the market today. However, both protocols have advantages and
disadvantages. IP switches and routers are ubiquitous and are the
preferred solution as a result of bandwidth efficiency. The new ATM
adaptation layer 2 (AAL2) standard now means that ATM switches
can be optimised for delay-sensitive speech and packet data services.
However, some infrastructure vendors advocate whatever solution
they have available. For example, Nokia, Nortel and Lucent
Technologies will advocate IP switches as a solution, while Ericsson,
Alcatel and Siemens will push ATM technology. It is worth noting that
Nortel and Lucent Technologies can offer both transport protocols.
Technically, there is no one solution which optimises the anticipated
demand in data traffic, if indeed there is to be a boom in the market.
However, the biggest offset is the cost of services offered as well as
the quality of service. Essentially, the success or failure of wireless
data will depend on selling the package to the end-user. Operators
will have to understand that their main position as a service and
network access provider has to be supported by other sectors such
as content provision and packaging, customer care and billing. They
have to decide what role they are to play in the emerging wireless
data market. As a result, Dataquest anticipates that network
operators will begin to seek partnerships to meet the changes in the
market dynamics.
David Tade is senior telecoms industry analyst at Dataquest.
|
