With the continuous development of AI, data center application scenarios are gradually divided into general-purpose...

The correct statements are B and C. Option B is true because general-purpose computing data center networks focus heavily on stability, reliability, service continuity, and flexible VM deployment. Technologies such as M-LAG improve gateway and uplink redundancy and support fast failover, while VXLAN enables large-scale Layer 2 extension for VM mobility and resource flexibility.

Option C is also true because intelligent computing networks, especially those serving AI training and distributed GPU clusters, require an extremely low-loss or lossless forwarding environment. Technologies such as PFC and ECN are standard methods used to build such networks and reduce congestion-related packet loss. Option A is not selected because modern general-purpose data center networks do not necessarily require the traditional core-aggregation-access three-layer architecture; contemporary data centers commonly use leaf-spine fabrics instead. Option D is not selected because although intelligent computing networks do require ultra-low latency and RoCE helps reduce network communication latency, the wording about “latency of data transmission within a compute node” is not the standard network-side requirement emphasized in HCIA-Datacom. The focus is mainly on inter-node network transport efficiency and lossless forwarding.