An Evolutionary Approach to Network Self-Organization and Resilient Data Diffusion

Data diffusion techniques enable a distributed system to replicate and propagate data across a potentially unreliable network in order to provide better data protection and availability. This paper presents a novel evolutionary computation approach to developing network construction algorithms and data diffusion strategies. The proposed approach combines a linear genetic program with a cellular automaton to evolve digital organisms (agents) capable of self-organizing into different types of networks and self-adapting to changes in their surrounding environment, such as link failures and node churn. We assess the effectiveness of the proposed approach by conducting several experiments that explore different network structures under different environmental conditions. The results suggest the combined methods are able to produce self-organizing and self-adaptive agents that construct networks and efficiently distribute data throughout the network, while balancing competing concerns, such as minimizing energy consumption and providing reliability.