Я тоже не понимаю.
Объяснишь или скажешь где почитать?
The variance in the queuing delay increases the tail latency for RDMA operations. As mentioned earlier, a 400 GbE port takes 80 nanoseconds to transmit a 4 KB frame. If the packets of an RDMA operation are delayed due to waiting in the queue behind 10 similar-sized packets, the completion time of this RDMA operation increases (degrades) by 800 nanoseconds.
https://www.cisco.com/c/en/us/td/docs/dcn/whitepapers/cisco-addressing-ai-ml-network-challenges.html
Читать с картинками можно тут
https://docs.nvidia.com/networking/display/mlnxofedv23070512/rdma+over+converged+ethernet+(roce)