Session Management in a 3G Network

Now compare this with a session-switched rich media exchange. As shown in Figure
14.3, in a session-switched exchange, a packet flow is established between two or more
users or two or more devices. The job of the application layer software is to increase
session persistency (that is, the length/character of the billable event) and session
complexity.
Referring to Figure 14.3, note that there is an initial small bandwidth exchange, then at
(a) an elementary packet stream is established. At (b) the session changes and dynamic
rate matching is used to track the amplitude of the offered traffic. The burstiness is
increasing as the session progresses. There could be a number of static matching step
functions where new supplemental channel streams are added or taken away. Any one
of these channel streams can be variable rate (dynamically matched). At (c), the session
is closed down and the whole billable event can be captured by the billing process.
Session persistency and session complexity together determine session value. The
complexity axis has variable amplitude value. This is a complex billable event. It is
potentially quite complicated to capture and represent the value in such an event—and
hence produce a billing rationale. We can simplify the billability by using visible quality
metrics—quality metrics that the user can experience directly, such as color depth,
frame rate, resolution, and audio quality. The session complexity does not or should
not need to be explained as part of the billing process.

Philosophically this is a connection-oriented exchange—a complex conversation
between two people and two devices, or multiple people and multiple devices. It
determines the amplitude (burstiness) of our offered traffic, which in turn determines
our overall bandwidth requirement.
To support bursty bandwidth, we either need to overdimension the size of the (wireless
or wireline) pipe to cope with peak loading or we need to buffer to smooth offered
traffic amplitude, which introduces delay and delay variability. This means consistency
has a cost: Consistent-quality complex rich media exchange is implicitly bandwidth
hungry, which means we then have to qualify whether we have enough
bandwidth available.
Table 14.3 compares wireline copper access, RF wireless access, infrared wireless
access, and optical access. Optical access gives us potentially 270,000 GHz of bandwidth.
Even just C Band and L Band gives us 10,000 GHz.
Bandwidth quantity and quality expectations increase over time and are driven by
copper access and optical access performance benchmarks. Application bandwidth
also increases over time, as does application dynamic range—the amplitude of the
information envelope. The more dynamic range we need to accommodate, the more
bandwidth we need.