Nowadays, by technology developments, software systems enlarge in scale and complexity. In large systems and to overcome complexity, software architecture has been considered as a connected notion with product quality and plays a crucial role in the quality of final system. The aim of the analysis of software architecture is to recognize potential risks and investigating qualitative needs of software design before the process of production and implementation. Achievement to this goal reduces the costs and improves the software quality. In this paper, a new approach is presented to evaluate performance of component-based software architecture for software systems with distributed architecture. In this approach, at first system is modeled as a Discrete Time Markov Chain and then the required parameters are taken from, to produce a Product Form Queueing Network. Limitations of source, like restrictions of the number of threads in a particular machine, are also regarded in the model. The prepared model is solved by the SHARPE software packages. As the result of the solution of the produced model in this approach, throughput and the average response time and bottlenecks in different workloads of system are predicted and some suggestions are presented to improve the system performance.