In this paper, we propose a graph-theoretic framework for designing architectures for distributed control. Specifically, the popular concept of community structure is used to decompose an integrated network into multiple sub-networks with minimum interactions. The state space of the network is represented as an equation graph (directed). Communities identified on this graph represent sub-controllers for the distributed control system. A quality measure 'interactivity' is defined to compare such decompositions. The proposed method has many advantages (e.g. possibility of non-square controllers, provision to ensure controllability and observability, scalability, etc.) over existing approaches. The effectiveness of the proposed framework is illustrated via several industrially relevant examples. (C) 2017 Elsevier Ltd. All rights reserved.