Traffic Calculation Methods
Like any mobile communications technology, traffic calculation and system
dimensioning for GSM begin with the estimation of how much traffic
demand there will be and from where it will come. In other words, one must
estimate the traffic demand in the coverage of each cell. This is rather an
inexact science. One can certainly acquire demographic data such as population
density, average household income, and so on. One can also acquire
data related to vehicular traffic in order to estimate traffic demand for cells
that cover roads. Based on these factors and others (such as how many competing
operators exist), one makes an estimate of the peak traffic demand
per cell. This estimate may well be incorrect. Fortunately, however, time is
an ally. In a new network, traffic demand grows gradually, which provides
the operator with sufficient time to monitor usage and more accurately predict
traffic demand over time.
Because all GSM traffic is circuit-switched, network dimensioning is a
relatively straightforward process once traffic demand per cell is specified.
The process largely involves determining the amount of traffic to be carried
in the busy hour and dimensioning the network according to Erlang tables.
The air interface, which represents the scarcest resource in the network,
is dimensioned with the highest blocking probability. Typically, network
designers dimension the air interface according to a two-percent blocking
probability (Erlang B). For a one-TRX cell with seven TCHs (BCCH, CCCH,
and SDCCH/4 are sharing timeslot 0), the cell can accommodate approximately
2.9 Erlangs. For a two-TRX cell with 14 TCHs (timeslot 0 on one carrier
is used for BCCH and CCCH and timeslot 1 is used for SDCCH/8), the
cell can accommodate approximately 8.2 Erlangs. For a three-TRX cell with
22 TCHs (one timeslot is allocated for SDCCH/8), the cell can accommodate
approximately 14.9 Erlangs. It is important to note that the traffic-carrying
capacity of each cell must be calculated independently.
Other interfaces in the network are usually dimensioned at much lower
blocking probabilities. For example, the A interface would typically be
designed for a 0.1-percent blocking probability. Similar blocking would
apply to other network-internal interfaces such as the interface between
the MSC and IWF. Typically, interfaces to other networks, such as the
PSTN, are dimensioned at slightly higher blocking probabilities—such as
0.5 percent. Of course, the choice of blocking probability for any interface is
a balance between cost and quality. The lower the blocking probability, the
higher the quality and the higher the cost. The higher the lower blocking
probability, the lower the quality and the lower the cost.
1466 times read