Practical Considerations in Trunk Engineering for Cellular Service

We identify and address practical problems facing the engineers who are responsible for trunk engineering(determining optimal trunk requirements between switching systems in a telecommunications network) in a nation-wide cellular service. Currently, Erlang B formula is used to calculate the number of trunks to carry the estimated cellular traffic with given target grade of service(i. e. call block rate). However, our recent measurement at a nation-wide cellular service covering more than 15 million customers shows that the measured block rate is occasionally far greater than the expected block rate, as much as 8 times. Fearing this, it is a common practice for field engineers to assign far more trunks than dictated by the Erlang B formula. But the main problem is that there is no basis on how to assign more trunks. In this paper, we track the cause for excessive block rate by analyzing vast amount of call log to identify the characteristic of the recent cellular traffic. We introduce a simple but effective compensation method to adjust the Erlang B formula with random and non-random traffic. The second problem we address is that the Erlang B formula gives average block rate while the management of the cellular service demands the engineers to guarantee given upper limit to the block rate. We employ the concept of the confidence interval to guarantee given block rate with certain reliability. We develop a simulation program to derive an updated version of Erlang B table with the confidence interval and a simple heuristic method to compensate for the peakedness of contemporary cellular traffic.