In this paper, the architecture of Digital Ecosystems Control based on Predictive Real-Time Situational models is presented. It has been reviewed Data Fusion issues in Digital Ecosystems and proposed a methodological approach for solving such issues based on modern big data technologies, investigated the Digital Ecosystem control in real-time as a situational management , modeled the architecture of the real-time DES control systems with in-memory technologies, message-oriented systems, virtualization, containerization and ability for horizontal scaling. Finally, the paper proposes the principles of predictive models developing for situational control. The methods are based on the intelligent analysis of DES statistical data and knowledge bases development. Retraining the model is carried out in real-time regime.