cutthrough

Cut-through, in the context of data networks, refers to a method of packet forwarding used by some network switches. In a cut-through switch, the switch begins forwarding a frame as soon as the destination address is read from the frame header, often before the entire frame has been received. This approach reduces end-to-end latency compared with store-and-forward switching, which waits until the whole frame is received and a CRC check is completed before forwarding.

Variants of cut-through switching include classic cut-through, which starts forwarding right after the destination address is known, and fragment-free cut-through, which forwards only after the first 64 bytes have been received to avoid propagating runt frames. In practice, many devices offer multiple forwarding modes or hybrids, allowing operators to balance latency against error checking and reliability.

The main advantage of cut-through is lower latency, making it attractive for latency-sensitive applications or environments where quick forwarding decisions are crucial. The main drawback is a higher risk of forwarding corrupted frames or runt frames, since the switch may begin transmission before validating the entire frame. Therefore, the suitability of cut-through depends on link quality, error rates, and the tolerance for potential frame errors. In modern networks, switching modes are often configurable, with cut-through used in specialized low-latency paths and store-and-forward favored for general-purpose, error-prone networks.