Performance evaluation of a dead reckoning mechanism

A dead reckoning mechanism allows one to reduce network bandwidth utilization considerably. As available bandwith is the key problem for large distributed interactive simulation (DIS) systems, we expect that a dead reckoning mechanism will allow us to increase the number of entities involved in a DIS exercise. In this paper, we model a dead reckoning mechanism and we present an evaluation of its performance. Such a mechanism reduces the number of program data units (PDUs) exchanged during the simulation, but it also completely changes the nature of the stochastic process which models the arrival of PDUs. Indeed, it adds some kind of sporadicity. We present some measurements on a DIS simulator which exhibits such a behaviour. Then we investigate the effects of such a mechanism in terms of response time and the number of allowed entities using a simulation tool. Finally, some extensions towards the numerical resolution of a large Markov chain associated with the problem are presented.