A model for flexible manufacturing cellular systems analysis has been introduced in this paper. Such a model could be generated directly or using higher level models such as stochastic Petri nets or discrete event simulation. A discrete-event system formulation and state partition into basic cells and fast and slow varying section, lead to a reduced computation cost. Further research in this area should focus on systems modeled with Markov chains which exhibit a cut-off phenomenon, as the existence of a cut-off phenomenon is a good indicator to whether a transient or a steady-state analysis is appropriate in a given setting. For example, if the cut-off time is known and the duration of observation is less than the cut-off time, then transient analysis is more meaningful than steady-state analysis. Identification and measurement of the bottleneck times in production lines has implications for both natures concerning the preventive maintenance and the production automation. In this paper we address the Markov model of production lines with bottlenecks. In lines where machines have identical efficiency, the machine with the smaller downtime is the bottleneck. In two-machine lines, the downtime bottleneck is the machine with the smallest value of p.Tup.Tdown, where p is the probability of blockage for the first machine and the probability of starvation for the second. Anticipation of events like full buffer or empty buffer, which determine bottlenecks, has also implications for the preventive maintenance of the manufacturing system. Future work in this area should focus on extensions of the results obtained in manufacturing systems with high failure rates.