By default, the globalSAN initiator will show the enabled policy "Header and Data (recommended)" for any connected targets.
This policy checks each transmitted SCSI command to ensure it's received correctly, and requests that it be sent again if a discrepancy is found.
The CRC32C (error detection) values for the HeaderDigest and DataDigest need to be in agreement on both the initiator and the target in order to be effective.
If there's a policy mismatch between the initiator and the target, globalSAN will show a warning when connecting.
'Header and Data' Error Detection setting was not applied on the target. This may potentially lead to data corruption. It is recommended to set the values of HeaderDigest and DataDigest iSCSI parameters on the target to 'CRC32C'.
The Error Detection policy does require some extra processing, and while this impact is often not noticeable using gigabit Ethernet, its effect can be especially apparent when using a higher bandwidth connection such as 10GbE.
Disabling the setting for 10GbE or higher connections will likely greatly improve throughput.
While it’s possible the error detection would result in slower throughput because it’s actually detecting and correcting errors, in most cases the throughput is slower simply because of the CRC algorithms themselves.
Though unlikely, it's also possible for an error to go unnoticed even with these cyclic redundancy checks in place.
A direct connection from the workstation to the server with no additional devices (such as switches or routers) will further reduce the chance of encountering any failed transmissions.
That said, maintaining current backups is always recommended, and more so in the event this protective policy is not enforced.