Mobile Network Architectures

The traditional network architecture used in GSM-MAP and ANSI 41 is very hierarchical—
a centralized mobile switch controller sits in the center of the network (see Figure
12.2). There may be a number of switches to cover a country. Each switch controls
a number of base station controllers, which in turn support the local population of
mobile users.

It is a conventional wireline network based on ISDN, but with a mobility management
overlay (Mobile Application Part). This provides the additional functionality needed
to move users from cell to cell (power control and handover); to set up, maintain, and
clear down mobile calls; and to bill for services provided to mobile users.
Going from left to right, the base stations talk to the BSC over the A-bis interface.
This interface takes the 9.6 kbps, 14.4 kbps, or 13 kbps voice traffic from the mobiles
(with some embedded signaling) and moves the traffic to and from the BSC over typically
multiplexed (120) 16 kbps traffic channels within a 2 Mbps pipe. In this example,
voice traffic is then transcoded from the 13 kbps (or 12.2 kbps EFR) codec stream to a
64 kbps low PCM (pulse coded modulated) channel stream by the transcoder (TCE).
An operation and maintenance center looks after hardware and channel monitoring.
At the middle of the diagram at the top we have the mobility management platform
consisting of the Authentication Register, the Visitor Location Register, the Home
Location Register, and the Equipment Identity Register. When a user is on his or her
home network in his or her home country, he or she is logged into the Home Location
Register. If the user travels to another country, the phone scans the 32 control channels
standardized in GSM and camps onto a serving network (this will either be the
strongest signal or a directed logon determined by inter network roaming agreements).
The visited network will log the user into the Visitor Location Register.

The Visitor Location Register needs to let the user’s home network know that the user
has moved. If someone now phones the user’s home number, the user’s call will be forwarded—
at some expense—to the user via the visited network. The authentication
register looks after SIM/U-SIM based user authentication and the equipment identity
register matches the user to the equipment being used. (Stolen equipment can be
barred from the network.) These mobility management functions involve, as you
would expect, substantial signaling, and this is carried over the SS7 signaling layer on
64 kbps multiplexed land lines.

The mobility management overlay provides the information needed for billing, so it
is arguably the most commercially important component in the network.
Traffic to and from mobile users is consolidated in the switch—hardware routed on
the basis of the target phone number used in the call setup procedure. If the call is
mobile to mobile, for example, the end-to-end link is determined by the sender’s IMSI
number and the receiver’s IMSI. When a call setup request is received at the MSC, the
MSC uses Layer 3 (network layer) signaling to allocate access network resources for
the call via a BSC and BTS. Layer 3 talks to Layer 2 (the data link layer) to allocate logical
channel resources via the BTS to the mobile. Layer 2 talks to Layer 1 to acquire
physical channel resources (that is, time slots within an RF channel in GSM/TDMA).
Figure 12.4 shows this layer modeling.
This all works fine when the traffic in both directions is more or less constant rate on
a per-user basis. Average call length in a cellular network is about 2 minutes. Traffic
loading can therefore be very accurately predicted. On the basis of these predictions,
decisions can be taken on how much backhaul bandwidth to install (how many 2 Mbps
lines to install).
Average call length is actually getting longer year by year as call rates reduce, and,
anecdotally, younger people also seem to take more than their parents’ share of time on
the phone. So call length is increasing as more young people start using mobile phones.
This nevertheless still represents quite predictable loading.
Historically, transmission bandwidth in the MSC and copper access network has
been overprovisioned to ensure that grade of service is more or less equivalent to fixed
access PSTN in terms of availability (so-called five 9s availability). You pick up the
phone, and 99.999 percent of the time, you get a line, or put another way, there is a 1 in
10,000 chance of the network being engaged. In practice, the limitation in a mobile network
tends to be the radio resource rather than network resources. One of the major
rationales of moving to a packet network, however, is to reduce the cost of network
transmission.

In a circuit-switched network, a logical channel and physical channel are established
end to end for the duration of the call. The logical channel and physical channel exists in
two directions simultaneously. Over the radio interface, the channel is a duplex spaced
RF channel pair 45 MHz or 190 MHz apart or (in TDD) an uplink and downlink time slot.
In a duplex voice conversation, we are only talking for approximately 35 percent of the
time; that is, for more than 50 percent of the time we are either listening to the other person
or pausing (to draw breath) between words. A pure packet-routed network avoids
this wasted bandwidth. Packets are only sent when voice activity is detected.
In defense of circuit-switched networks, it is valid to point out that there is a fundamental
difference between logical channel allocation and physical channel allocation.
Over the radio air interface, a logical channel pair will have been allocated for the duration
of a duplex voice call. However, if the handset and the base station are using discontinuous
transmission (RF power is only generated when voice activity is detected),
then there is no physical occupancy of the radio layer.
Similarly, because much of the core transmission network has been historically overprovisioned
and, in many cases, fully amortized, increasing core network bandwidth utilization
is neither necessary nor cost-effective. We do need to take into account, however,
the increasingly bursty nature of the traffic being offered to the network; that is, we are
justifying the transition to packet networks on the basis of their suitability for preserving
the properties of bursty bandwidth—a quality rather than cost-saving justification.
It is as problematic as it is difficult to put a finite value on quality; how much is a 24-
bit color depth 15 frame per second video stream worth compared to a 16-bit 12 frame
per second video stream. Additionally, we need to factor in the extra costs incurred by
deploying packet routing in the network. Figure 12.5 shows the first changes that have
to be made—the addition of a GPRS or packet traffic support node.

The GPRS support node talks to the BSC across a 2 Mbps (2.048 Mbps) ATM transport
layer. (We case study ATM in Part IV of the book.) For the moment, all we need to
know is that the ATM layer allows us to multiplex bursty traffic and maintain its time
domain properties. What you put in at one end of the pipe comes out at the other end
of the pipe unaltered—a bit like a filter with a constant group delay characteristic. The
traffic experiences some delay because of the multiplexing—and some delay along the
transmission path—but the delay is a constant and is equal for all offered traffic. At
either end of the ATM pipe, we can, of course, buffer traffic and prioritize access in the
ATM pipe. That is, traffic is all treated equally while it is inside the pipe but can be
given differential transport priority before it gets into the pipe.
The example shown in Figure 12.5 highlights new MAC (Medium Access Control)
functionality (Layer 2 functionality) in the mobile and BSC. This is to support highspeed
circuit-switched data from a handset capable of using more than one time slot on
the uplink and downlink; that is, variable bit rate can be delivered in increments of
additional time slots (or in IS95, additional PN offsets). The A-bis and UM/IF interface,
therefore, remains essentially unchanged.
Figure 12.6 shows the addition of a serving GSN that can manage simultaneous
circuit-switch and packet-routed traffic talking to the gateway GSN using IPv6 (case
studied later). The SGSN talks to the BSC over an ATM transport layer. The BSC talks
to the BTS (also over ATM), and the BTS exchanges packets with the mobile using EGPRS
radio blocks to manage packet re-sends (covered earlier in Chapter 2, where we
discussed system planning). It is interesting to note the continuing presence of an MSC
and an interworking function to manage simultaneous packet-routed and circuitswitched
traffic.

The forward compatibility selling point is that once an operator has put ATM
in between the SGSN and BSC and BTS, it is then relatively easy to implement
IMT2000DS with a MAC layer delivering dynamic rate matching on a 10-ms frame resolution,
effectively wireless ATM. IPV6 can be used to provide some higher-layer prioritization
of packet streams, establishing rights of access and priority/preemption
entitlements to network and radio transmission bandwidth.
There is still (particularly in Europe and Asia) a strong circuit-switched feel to the
network. ATM is a hardware-based implementation of virtual circuit switching.
Remember also that the genesis of the 3GPP specification was to implement wireless
ISDN over the radio physical layer. 289