Typical 3G Network Architecture

Shows the major components in a 3G network, often described as an IP QoS
network—an IP network capable of delivering differentiated quality of service. To
achieve this objective, the IP QoS network needs to integrate radio physical layer performance
and network layer performance. The IP QoS network consists of the Internet
Protocol Radio Access Network (IPRAN) and the IP core replacing the legacy Mobile
Radio Switch Center (MSC).
The function of the Node B, which has replaced the base station controller (BTS), is
to demodulate however many users it can see, in RF terms, including demodulating
multiple per-user traffic streams. Any one of these channel streams can be variable bit
rate and have particular, unique QoS requirements. Node Bs have to decide how much
traffic they can manage on the uplink, and this is done on the basis of interference
measurements—effectively the noise floor of the composite uplink radio channel.
A similar decision has to be made as to how downlink RF bandwidth is allocated.
The Node B also must arbitrate between users, some of whom may have priority access
status. We refer to this process as IPRAN interference-based admission control.

The RNCs job is to consolidate traffic coming from the Node Bs under its control.
The RNC also has to load balance—that is, move traffic away from overloaded Node
Bs onto more lightly loaded Node Bs and to manage soft handover—a condition in
which more than one Node B is serving an individual user on the radio physical layer.
The fourth handset down on the right-hand side of Figure 11.1, for example, is in soft
handover between two Node Bs supported by two different RNCs. The handset is,
more or less, halfway between two Node Bs. To improve uplink and downlink signal
level, the RNC has decided that the handset will be served by two downlinks, one from
each Node B. Both nodes will also be receiving an uplink from the handset. Effectively
this means there will be two long codes on the downlink and two long codes on the
uplink. The two uplinks will be combined together by the serving RNC, but this will
require the serving RNC to talk to the other serving RNC (called the drift RNC). The
same process takes place on the downlink.
The RNC has to make a large number of very fast decisions (we revisit RNC software
in Chapter 17 in our section on network software), and the RNC-to-RNC communication
has to be robust and well managed. The RNCs then have to consolidate
traffic and move the traffic into the IP core. Admission control at this point is done on
the basis of congestion measurements:
 Is transmission bandwidth available?
 If no transmission bandwidth is available, is there any buffer bandwidth
available?

If no transmission bandwidth is available and no buffer bandwidth is available then
packet loss will occur unless you have predicted the problem and choked back the
source of the traffic.
RNC traffic management and inter-RNC communication is probably the most complex
part of the IPRAN and is the basis for substantial vendor-specific network performance
optimization. This is complex deterministic software executing complex decisions within
tightly defined timescales. As with handset design, there is considerable scope for hardware
coprocessors and parallel hardware multitasking in the RNC. As with handset
design, we will show that network performance is also dependent on the RF performance
available from the Node B—the Node B’s ability to collect highly bursty traffic from users
and deliver highly bursty traffic to users. 234