Generation and Solution of Markov Chains Using MOSES

In this paper the Markov analyzer MOSES (MOdelling, Speci cation and Evaluation System) and the model description language MOSLANG { both developed at the Institute for Operating Systems at the University of Erlangen{Nuernberg { are described using two examples. To evaluate the performance of a computer system the system has to be speci ed. In this paper the model description language MOSLANG is introduced and applied to some examples. The core of MOSLANG consists of constructs suitable for the speci cation of the possible states of the system and of RULE constructs which model the state transitions. This speci cation method is much more compact than other comparable methods and enables the user to specify large systems when he has become familiar with MOSLANG. The Markov analyzer MOSES supports the input of MOSLANG and subsequently creates the Markovian system of equations automatically. For solving this system of equations ve di erent methods are provided. MOSES calculates the state probabilities and derives from them the performance measures which are speci ed using MOSLANG. MOSES has been used to solve many di erent problems, among others for analyzing and comparing di erent multiprocessor versions of the operating system UNIX. MOSES is implemented in the programming language C and runs on all UNIX systems.