SCADA solutions are under a high flux to shift their focus from process control of a limited set of industrial plants to the control of large-scale system of systems. This is in par with the recent proliferation of ubiquitous/pervasive computing paradigm mostly embodied as Internet of Things (IoT). In a traditional setup, a whole system is only partially covered by SCADA data points; therefore, complex simulation is required to fit the missing measurements, hence, buttress decision support scenarios. This usually entails a fully integrated and centralized approach, where SCADA infrastructure needs to hold and distribute data, both collected and calculated. It leads to the increase of load on a supporting real-time database, which hosts millions of data points. It is a challenge that a traditional SCADA design (based on a shared memory database and competing processes), cannot fulfill in real-time. This paper proposes an alternative approach of an architecture and basic functionality of a SCADA system. The proposed architecture targets distributed SCADA systems that can be used to supervise and control large-scale distributed industrial or infrastructural systems. Strategic data organization and segmentation are introduced, so that the acquired data can be efficiently distributed throughout the system. The proposed architecture pushes forward a peer-to-peer node structuring scheme, where an autonomous node supervises and controls only subsets of the system. Nodes collaborate to establish a unified view of the entire system. The proof of the concept implementation has proven to be able to manage significantly more data points in a distributed fashion than a centralized variant.