STAMINA will leverage statistical physics inspired methods to deliver a novel foundational framework for managing complexity in information network mega-structures and for efficientlly solving large-scale network optimization problems that are intractable by classical methods.Recent success stories like the effective decoding of LDPC codes in information theory support the great promise of the approach.A cross-disciplinary work plan is proposed, at the interface of statistical physics, networking and computer science.A network optimization problem with given objective over a space of possible configurations is mapped to a statistical physics problem instance with probability distribution over possible configurations.Solving the optimization problem is equivalent to finding minimum energy configurations where probability distribution concentrates.Statistical mechanics theories from spin glass and disordered systems will establish fundamental connections among atomic micro-interactions, emergent network behaviour and phase transitions.Belief propagation message passing methods will be harnessed, that disassemble the hard centralized combinatorial problem to iterative lightweight local messaging, thus achieving autonomic network control at no cost for solution dissemination, and promoting green computing through ultra-low processing load.Flexibility and simplicity enable real-time adaptation at different time scales of variations through online construction of solutions.Three challenging case studies (energy-prudent control at device and network level, resource management regimes for optimal transport capacity and latency, and inference of hidden network states) serve as proof-of-concept for enabling novel, currently suppressed functionalities.A solid validation plan is laid, with large-scale simulation and test-bed experimentation. Notable achievements of members of our team make us optimistic about the potential of the methods and motivate our research agenda.