To develop an efficient parallel application is not an easy task. Applications rarely achieve a good performance immediately therefore, a careful performance analysis and optimization are crucial. These tasks are difficult and require a thorough understanding of the program’s behavior. In this paper, we propose an on-line performance modeling technique, which enables the automated discovery of causal execution flows, composed of communication and computational activities, in MPI parallel programs. Our model reflects an application behavior and is made up of elements correlated with high-level program structures, such as loops and communication operations. Moreover, our approach enables an assortment of on-line diagnosis techniques which may further automate the performance understanding process.