The paper presents a methodology of assessment of distributed systems in the presence of exceptions and errors. The analysis is done by identifying data transitions for a system and deriving general characteristics about system behavior. Additionally, characteristics of internal subsystems are calculated. Distributed dependencies are modeled as basic dependencies between methods and compound dependencies existing in the system. The analysis answers the question how distributed interactions influence the result of execution, which supports the identification of correct programming constructs. The theoretical model is used in a developed tool that allows to automatically modify, trace and analyze distributed applications. Additionally, a simulator is developed to estimate the behavior of a real system. An example of experimental use of the methodology is presented.