The paper presents a proposal of a hybrid method for embedded system modeling with emphasis on constraints analysis. Since one of the major factors for most embedded system from the performance point of view is the real-time response of component execution the proposed method focuses on analysis of attributes affecting such performance requirements and optimizes the design process accordingly. The core idea of the method is to separate components into three distinctive groups as critical, hybrid and non-critical and select an appropriate memory and resources sharing model accordingly. Experimental case study is provided to verify the method performance compared to other available modeling methods like best-effort or dependable embedded system design methods. The hybrid method shows its advantages in more effective use of available implementation platform resources while meeting all of the required real-time response constraints.