Hi Alan,
When the mtu is set to 1500 throughout your network, there is little risk of generating frames with a size in excess of 1536 bytes.
The payload is 1500 bytes, that is your ip header and data. Ethernet adds 14 bytes to that plus an extra four at the end. The latter contain the CRC checksum which is needed to error-check the frame. So the entire frame is then 1518 bytes long. This is true for a standard IEEE ethernet frame.
When you are using a trunk (dot1Q) this adds an extra four byte tag to the frame. Note that this makes the frame larger than the IEEE specification.
Should you be utilizing QinQ, there will be another four bytes added to that.
When you do the math, it adds up to 1518 +4 + 4 = 1526. Still within a 10-byte safety margin from where the mux will crash.
It is also true that Cisco switches will forward giants, frames with a size in excess of 1518 bytes.
The reason for this is mainly the requirement to support dot1Q but as you have seen, it is not likely that a dot1Q tagged frame will become too large.
In the way you presented the case, I cannot see any risk of giants being sent to the mux.
The biggest risk is the transmission guy got his numbers wrong and the critical value is < 1522.
regards,
Leo
