What is the difference between store-and-forward and cut-through switching?

These are two internal packet processing methods that trade off latency vs error detection.

Store-and-Forward

• Switch buffers entire frame before forwarding
• Validates CRC (checksum) — drops errored frames
• Higher latency (depends on frame size)
• Standard in virtually all modern switches

Cut-Through

• Starts forwarding as soon as destination MAC is read
• Only ~64 bytes buffered
• Lower latency (typically 200-500ns vs 5-10μs)
• Cannot detect errors — forwards bad frames

Fragment-Free (Modified Cut-Through)

• Waits for first 64 bytes (collision detection window)
• Filters most malformed frames
• Compromise between the two

Modern Switch Behavior

• Most switches use store-and-forward by default
• Some support "adaptive" — cut-through until errors detected
• Data center switches often offer cut-through for HFT (high-frequency trading)

Where Cut-Through Matters

• High-frequency trading (nanoseconds count)
• Some real-time industrial protocols
• Storage networks in some designs

Where Store-and-Forward Wins

• General enterprise networks
• Networks with mixed speeds (must buffer for speed conversion)
• Wherever error detection matters (most cases)
• Long-distance and lossy links

Speed Conversion

When forwarding from a fast port to a slow port (10G → 1G), store-and-forward is mandatory. Cut-through only works between same-speed ports.

Verdict

For 99% of applications, store-and-forward is the right choice. Cut-through is a niche optimization for specialized low-latency scenarios.