基于节点分离双路径的高性能网络传输时长优化OA

We address the problem of minimizing data transfer duration for bandwidth reservation requests in high-performance networks by proposing a novel scheduling approach based on two node-disjoint paths.By integrating path selection,bandwidth allocation,and time window optimization,we formulate a mathematical model aimed at minimizing transfer duration and rigorously prove that the problem is NP-complete.To tackle this computational challenge,a greedy heuristic algorithm is designed,which employs global time dot ordering,node-disjoint widest-path search,and early termination mechanisms to achieve near-optimal solutions in polynomial time.The proposed algorithm first integrates the time states of all links,generates a globally ordered sequence of time points,and builds candidate time windows based on this.Subsequently,within each window,two node disjoint paths with the maximum total bandwidth are searched in parallel,and greedy scheduling is performed in the order of increasing window length.Extensive simulations conducted on a topology emulating the ESnet infrastructure demonstrate that compared with the conventional single-path scheme Opt-SD,the proposed algorithm reduces the average transfer duration by 16.65%,increases the scheduling success ratio by 21.62%,and improves processing time per request by 63.01%.The results indicate that the proposed algorithm effectively balances solution quality and computational efficiency,offering a practical and scalable solution for large-scale data transfer scheduling in high-performance network.