The Hamiltonian Property of the Data Center Network DPCell
Dong Hui, Huaqun Wang, Mengjie Lv, Weibei Fan · IEEE Transactions on Network Science and Engineering · 2025
Data center networks (DCNs) play an irreplaceable role in digital transformation by enabling data storage, processing, and complex computing. DPCell is a DCN built on a novel fabric of switches structure, outperforming other DCNs based on dual-port servers in scalability and bisection width. Ensuring reliable communication performance in DCNs is critical for continuous service delivery. In this paper, we investigate the reliable communication performance of DPCell from the perspective of Hamiltonian properties. We first prove that DPCell is Hamiltonian-connected. Considering the inevitability of network failures, we further prove that DPCell is a super fault-tolerant Hamiltonian network under a fault model, achieving an optimal result. The superior Hamiltonian properties of DPCell confirm its reliable communication performance, laying the foundation for deadlock-free reliable communication and fast adaptive diagnosis. We verify the theoretical results through simulation experiments and further evaluate DPCell's Hamiltonian properties under conditions exceeding the fault model limit, including structure faults caused by network attacks.