Performance evaluation of dynamic priority operating systems

Operating systems which implement a dynamic priority mechanism are very common. Nevertheless, it is very difficult to develop an accurate analytical model to evaluate their performance, mainly due to the different forms of dependency between the various constituent parts. We introduce a black box modeling approach which allows us to identify and decompose the different functions of the operating system into elementary subsystems from which submodels are developed. Interactions between submodels are taken into account while developing the system model. We then apply this modeling technique to the study of operating systems with dynamic priorities. In particular, we investigate the performance of the BS2000 operating system, which implements a dynamic priority mechanism, by means of a stochastic reward net (SRN) model. The results are then compared against an approximate model in which a static priority mechanism is assumed. Real system measurements are also carried out in order to validate the analytical results