An efficient and reliable service customized routing mechanism based on deep learning in IPv6 network

Best-effort service model of traditional routing is gradually hard to meet the personalized demands under rapid development network technologies (e.g. 5G and IPv6). Therefore, customization should be considered. In this work, a customized mechanism based on deep learning in IPv6 proposed, which includes learning-based module, reliability evaluation calculation module. The first module uses neural learn complex function, can quickly output win-win strategies user demands. second quantify paths, where not only link status Neighbor Unreachable Detection (NUD) considered, but also propose performance weights ensure differentiated performance. third gray wolf optimization algorithm calculate an optimal path forward services with as constraints maximum minimum cost goal. Finally, tested Source Address Validation Improvement (SAVI) platform, reduce execution time by 12.25% improve average reliability, ISP satisfaction 9.0%, 40.45% 7.4%, respectively.