Remote Direct Memory Access: application → Queue Pair → Memory Region → RNIC → fabric → remote memory, with kernel & CPU bypass. Pick an operation and step through it.
Speed
1×
0
Ready
Choose an operation above and press Play, or step through with Next.
Key idea: once memory is registered, the RNIC moves data straight between registered buffers by DMA.
The kernel is off the data path (kernel bypass), and for one-sided WRITE/READ the remote CPU is never interrupted (CPU bypass, zero-copy).
←/→ step · Space play/pause.