弹性光网络中面向提前预约服务请求的生存性感知路由频谱分配算法OA

Multi-path protection can effectively enhance the service survivability of light-path requests by transmitting data through multiple disjoint paths.This approach ensures that even if one path fails,the service remains uninterrupted.This paper focuses on the survivability-aware routing,modulation,and spectrum assignment(RMSA)problem in elastic optical network(EON)based on deep reinforcement learning(DRL)for advanced reservation(AR)service requests,where AR requests not only allow for im-mediate allocation of network resources upon arrival but also enable scheduling at a future time,character-ized by advanced reservation and waiting features.To efficiently determine the multi-path routing,modu-lation format,spectrum allocation,and scheduling time for AR service requests,this paper proposes a dy-namic resource allocation algorithm based on DRL.Simulation results on a 14-node network topology demonstrate that the proposed DRL algorithm reduces the blocking probability by as much as 31.9%at 300 Erlangs traffic load.