Cut-Through Switching and CRC Stomped [Explained]
What is Cut-Through Switching
Cut-through switching refers to a type of switching mode that allows the switch to begin forwarding a packet as soon as the destination MAC address is known, rather than waiting for the entire packet to be received before forwarding it.
In cut-through switching mode, the switch reads the destination MAC address from the incoming packet and immediately forwards it out the appropriate port, without buffering the entire packet. This can result in lower latency and faster packet forwarding, but it also means that errors in the packet may be forwarded along with it.
There are several different types of switching modes in Cisco Nexus switches, including store-and-forward switching, where the switch buffers the entire packet before forwarding it, and fragment-free switching, which buffers only the first 64 bytes of the packet before forwarding it.
The choice of switching mode depends on the specific requirements of the network and the applications running on it. Cut-through switching is often used in high-performance environments where low latency is critical, such as in data center and cloud computing networks.
What is Cut-through Switching
Server-1 is sending packets to Server-2, the packets are corrupted.
On Switch-1:
- Switch-1 receives the frame on eth1/10 and decides that it needs to send out the packets on eth1/20. After the FCS (Frame Check Sequence), the Nexus will know that the packet is corrupted, so it will increment the CRC on ingress eth1/10 (which means the server facing interface sent a corrupted frame)
- On the outgoing interface, it will change the CRC-check value with the STOMPED value, and forward the rest of the frame. At the same time, it will increment CRC_STOMPED on egress interface eth1/20.
On Switch-2:
- Switch-2 receives the frame on eth1/1, and decides that needs to be send out to eth1/2. Will receive the trailer and know that this frame has CRC_STOMPED value (which means Switch-1 didn’t generate the corrupted, just forwarded it because he is a cut throw switch). So it increment CRC_STOMPED counter on both ingress interface eth1/1 and egress interface eth1/2.
Troubleshoot Notes:
- Because there are problems for Server-1 and Server-2 communication, you will need to troubleshoot this problem. You will start with SW-B, will look on eth1/2 and see that hey, we have CRC_STOMPED packets on this interface. You will also look on eth1/1 (incoming interface) and you will see the same CRC_STOMPED packets.
- At this point, you will know that the CRC packets were not generated by Switch-1, because they are received stomped. So what you will do as a next point, you will look for incoming interface on Switch-1.
- You check the Switch-1 incoming interface eth1/10 and find that on this interface only CRC counter is increasing. This means that the frames are corrupted on this link between Switch-1 and Server-1
Reference: Sergiu.Daniluk
