Distributed Physical Device Connection Relationship Discovery Technology Based on Traffic Information

Cloud data center networks have a complex structure, in which multiple network protocols are utilised to participate in the composition of the data center. How to achieve topology discovery and mapping in the face of complex, multi-structured cloud data centre network devices has become a hot research problem. Network tomography (NT), as an end-to-end measurement technique of network boundaries, enables protocol-independent network measurements without the collaboration of internal nodes. The emergence of this technique provides a new research idea for cloud data center network topology mapping. In this paper, we propose a topology discovery technique based on the delay covariance matrix and a topology optimisation method based on the delay peak map, and conduct simulations for these two algorithms to verify the feasibility of the protocol-independent topology discovery technique applied to data center topology discovery.