Abstract: By introducing the digital twin network technology into the computing power network, the virtual mapping network of the computing power network can be established. Digital twin network system realizes efficient analysis, diagnosis, and control of computing power network through high-fidelity real-time interaction between virtual and real. Taking the self-optimization of computing power network as an example, a structural self-optimization model of digital twin computing power network is proposed, which realizes the real-time closed-loop control of network self-learning, self-verification, and self-evolution in digital twin computing power network. Different from the traditional self-organizing network (SON), the physical network infrastructure is separated from the SON module, and the SON self-optimization process is migrated to the virtual network, which reduces the complexity of computing power network operation and maintenance, and improves the flexibility and adaptability of the network. Simulation results show that the introduction of digital Siamese network technology can quickly deal with the problem of computing power network service timeout and reduce the overall service delay of the network.
Keywords: digital twin; computational network; SON; genetic algorithm