Recent years has seen profound changes in building technologies both in Europe and worldwide. With the emergence of Smart Grid and Smart City concepts, the Smart Building has attracted considerable attention and rapid development. The introduction of novel information and communication technologies (ICT) enables an optimized resource utilization while improving the building performance and occupants' satisfaction over a broad spectrum of operations. However, literature and industry have drawn attention to certain barriers and challenges that inhibit its universal adoption. The Smart Building is a cyber-physical system, which as a whole is more than the sum of its parts. The heterogeneous combination of systems, processes, and practices requires a multidisciplinary research. This work proposes and validates a systems engineering approach to the investigation of the identified challenges and the development of a viable architecture for the future Smart Building. Firstly, a data model for the building management system (BMS) enables a semantic abstraction of both the ICT and the building construction. A high-level application programming interface (API) facilitates the creation of generic management algorithms and external applications, independent from each Smart Building instance, promoting the intelligence portability and lowering the cost. Moreover, the proposed architecture ensures the scalability regardless of the occupant activities and the complexity of the optimization algorithms. Secondly, a real-time message-oriented middleware, as a distributed embedded architecture within the building, empowers the interoperability of the ICT devices and networks and their integration into the BMS. The middleware scales to any building construction regardless of the devices' performance and connectivity limitations, while a secure architecture ensures the integrity of data and operations. An extensive performance and energy efficiency study validates the proposed design. A "building-in-the-loop" emulation system, based on discrete-event simulation, virtualizes the Smart Building elements (e.g., loads, storage, generation, sensors, actuators, users, etc.). The high integration with the message-oriented middleware keeps the BMS agnostic to the virtual nature of the emulated instances. Its cooperative multitasking and immerse parallelism allow the concurrent emulation of hundreds of elements in real time. The virtualization facilitates the development of energy management strategies and financial viability studies on the exact building and occupant activities without a prior investment in the necessary infrastructure. This work concludes with a holistic system evaluation using a case study of a university building as a practical retrofitting estimation. It illustrates the system deployment, and highlights how a currently under development energy management system utilizes the BMS and its data analytics for demand-side management applications.