This paper presents the design, implementation, and the first experimental results of an architecture for autonomous demand-side load management (ADSM) system for Smart Buildings in view of influencing the energy demand in the Smart Grid. In such an architecture, the controller has a layered structure and appliances' operation is modeled as a finite state machine, which enables an efficient load control using scheduling techniques borrowed from real-time computing systems. In this study the Admission Control, which is the bottom layer of the presented architecture interacting in real-time with physical equipments, is addressed and the real-time power consumption management in a residential dwelling is implemented and tested in a real offce building. The experimental results provide a proof of concept for the proposed architecture and demonstrate the applicability of the online scheduling approach for ADSM systems for Smart Grids.