Performance modelling of automatic call distributors
Mark Perry · 1991
The purpose of the research has been to formally describe a sophisticated Automatic Call Distributor (ACD) and model its performance. The research may be used to provide guidelines on (1) provisioning operators, (2) deflecting calls, (3) prioritizing calls, and (4) turning call waiting lamps off and on. The formal description of a sophisticated ACD provided the necessary detail to develop performance models. The modelling used both queueing theory and simulation. The research was motivated by the fact that existing ACD modelling lagged behind the development of new sophisticated ACDs. Up until the last few years, most ACDs have had a simple and inflexible way to distribute calls to agents: the call that had waited the longest was given to the agent who had been ideal the longest. The modelling for these simple systems used the well known M/M/m queueing model. However during the past few years a new generation of ACDs have been developed which allow for very flexible call distribution. That is they can be configured to prioritize different types of calls and favor the selection of different types of agents. But while ACDs have grown more sophisticated, the performance modelling of them has not. For the most part, the M/M/m model is still used. This thesis develops more sophisticated models through priority queueing, simulation and random walks and applies them to the ACD application operator services.