An evolutive framework for server placement optimization to digital twin networks

Digital twin network (DTN) is a foremost enabler for efficient optimization in modern networks, as it owns massive real-time data and requires interaction with the physical network in real-time. When constructing a DTN, it is necessary to deploy many servers in the physical network for digital models' storage, calculation, and communication. Evolutionary algorithms show outstanding global optimization capabilities compared to the constructive heuristic method in such an optimization problem. However, due to the high dimensionality of the problem and the complicated evaluation of the deployment plan, evolutionary algorithms easily fall into the optimum local at a high computational cost, given that the server placement problem is an NP-hard combinatorial optimization problem. In this research, we propose an evolutionary framework for server layout optimization that significantly improves the optimization efficiency of evolutionary algorithms and reduces the algorithm's computational cost. An offline-learning-based approach is used to reduce the search space, and a self-examining guided local search method is proposed to improve the search efficiency. Additionally, a look-up table-based hybrid approach is used for solution evaluation, reducing computational overhead. Experimental results show that the proposed framework and optimization strategy can significantly improve the evolutionary algorithm search efficiency and achieve excellent convergence performance.