The forecast demand in the market for a service is formulated, first in terms of the number of customers or connections, as a market penetration, and then the average load or traffic per customer or connection. Not all of the traffic inputs are required, depending on the type of service and the preferred formulation for calculating the results. The Distribution parameters allow the peak traffic to be inferred from the traffic volume, or vice versa; these contrasting approaches to the calculations are common choices for voice and data services, respectively. The pivotal Traffic Calculation input (see 10.3.8.3 Distribution below) determines which inputs will drive the calculation, and which will be ignored, as follows:
Volume Driven: the Busy Hour Traffic result is inferred from the (average) Traffic per Connection input, according to the stated Distribution parameters, and the Nominal Bandwidth per Connection and Contention Ratio inputs are ignored
Peak Driven: the Busy Hour Traffic result is calculated directly from the Nominal Bandwidth and Contention Ratio inputs; the Annual Traffic result is then inferred from the busy hour traffic, according to the stated Distribution parameters, and the Traffic per Connection input is ignored
Independent: used for some media services for which the peak traffic and traffic volume can be estimated directly without reference to the Distribution inputs, which are ignored in this case.
The busy-hour traffic for a voice service is traditionally measured in terms of the average calling density in the busy hour (commonly measured in Erlangs). This is naturally inferred from the (average) Traffic per Connection input according to the spread of traffic over the year, and also between the busy period (hour) and the rest of the day. With this default Volume Driven calculation, the Nominal Bandwidth per Connection and Contention Ratio inputs (which are mostly intended for data services) are ignored.
A separate Erlang Transformation may be used to relate randomly-arriving busy-hour traffic to required capacity, depending on the number of discrete aggregation Sites and desired Grade of Service – see 10.3.10 Erlang B formula.
For a data service, the Traffic Calculation input can be set to Peak Driven, meaning that the peak traffic is calculated directly from the Nominal Bandwidth and Contention Ratio inputs (e.g., 1 Gbit/s, overbooked to some degree). In this case the traffic volume (e.g., in TBytes) if required is then inferred from the busy hour traffic, according to the stated Distribution parameters, and the Traffic per Connection input is ignored. (You would set the input Annual to Busy-Hour Unit Ratio = 60 * 60 / 8 / 1024 in this example to convert from instantaneous Gbit/s to cumulative TBytes in an hour.)
Although there is a lack of consensus as to a standard parameterisation of a data service, the STEM parameterisation is flexible enough to form the basis for any type of data service. The implementation is based on the common requirement for the demand to be peak driven, and provides inputs for the standard parameters of nominal bandwidth and contention ratio. Users may have particular parameters that they wish to use for specific data services, such as burstiness, or duty cycles. However, these are basically internal to the services concerned and can be readily captured as User Data.
For a media service, the peak and traffic volume assumptions may both be better estimated directly and the Traffic Calculation input set to Independent. This means that the Distribution parameters are ignored, which may be preferable if the day-to-day traffic distribution is unpredictable. For example, annual traffic volume may be constrained independently by spending.