One of our customers owns 2 AMD Express which are feeded with network flows coming from a Gigamon switch which has 'TCP Slicing' feature enabled in order to limit bandwidth consumption.
Since 'TCP Slicing' feature has been enabled, capacity reports show around 40%-50% of 'bad checksum packets' errors.
I believed it was possible to add a specific parameter (net.chcksum=false) within rtm.config as a workaround but the customer tried it and it has not change values from capacity reports.
Thus, the question is how to exclude checksum errors checks from errors scope or how to manage sliced TCP packets on AMD Express ?
Thanks for your help,
The AMD must receive full packets with correct checksums. The TCP slicing feature causes the AMD to be unable to calculate the correct byte counts (as the real packet length is lost), so in addition to problems related to traffic decoding mentioned by Chris, metrics like bytes, traffic volume or realized bandwidth are also significantly skewed.