Why would you choose store-and-forward switching over cut-through?
Because it reads the whole frame first, store-and-forward can drop corrupt frames before they spread, and that same full-frame buffering is what makes QoS classification and speed-mismatched ports possible — at the cost of slightly higher latency.
The single idea behind store-and-forward is that the switch buffers the entire frame before sending it, and three useful capabilities fall out of that one behaviour:
- Error detection. With the whole frame in hand the switch can recompute the CRC (Cyclic Redundancy Check) and compare it to the FCS (Frame Check Sequence). A corrupt frame is discarded instead of forwarded, so it doesn't waste downstream bandwidth — something cut-through can't do because it forwards before the FCS even arrives.
- Required for QoS (Quality of Service). Classifying and prioritising traffic (e.g. putting VoIP ahead of web downloads) means inspecting fields throughout the frame, which is only possible once the full frame is buffered.
- Required for ports at different speeds. If a frame comes in at 1 Gbps but must leave a 100 Mbps port (or vice versa), it has to be held in a buffer to bridge the rate difference — exactly what store-and-forward already does.
Trade-off: because it waits for the last bit before forwarding, latency is higher than cut-through (which starts forwarding after reading the destination MAC). For most networks the integrity and flexibility are worth it — store-and-forward is the default on Cisco Catalyst switches.
Go deeper:
Cut-through switching (Wikipedia) — explains why only store-and-forward can drop corrupt frames, since cut-through forwards before the FCS arrives.