Why does the benchmark use only two, equally likely, rates?
• The choice of two rates means that a realistic spread of rate precisions (about 3 bits) is used in the calculations, giving a reasonable estimate of the computational effect of multiple rates. The 50% choice is roughly the proportion between low-rate and high-rate calls in countries with a high penetration of mobile phone usage. The 50% ratio also means that the split can be determined fairly and cheaply from the bottom (least significant) bit of the number n, regardless of whether the number is expressed in a decimal or binary base. This allows the data file to be a simple ‘generic random number distribution’ instead of being a data file specific to this benchmark. • Only two tax rates are used. Billing schemes can be far more complex, quite often with more than two tax authorities involved. Also there are sometimes special plans which allow free calls up to a fixed monthly fee, and so on. Why not add more calculations into the benchmark? • Billing calculations do vary considerably.
