Research on End-to-end Forwarding Path Selection Method for CFN

With the rapid development of AI technology, computational resource-centered network becomes one of the main evolutionary directions of SDN network. Compute-first network (CFN) takes compute resources as the primary reference basis for network forwarding path selection, so how to make path selection based on the demand of compute resources required by customer services is one of the main research problems of CFN. In this paper, based on the network architecture of CFN, we analyze the selection and decision-making process of traffic forwarding paths from edge networks to computing centers in CFN according to the actual enterprise business deployment requirements, and combine the technologies of SD-WAN, SRv6, etc., to realize an end-to-end forwarding path selection model in compute-prioritized NICs, and put forward an end-to-end forwarding path quality detection method. This paper solves the problem that the end-to-end forwarding paths do not meet the actual service computing resource requirements because of the inconsistency of the forwarding path selection process of different sub-networks in CFNs. The experimental results prove that the end-to-end network forwarding paths selected according to the method in this paper better meet the customer’s application service requirements.