Understand rbd diff_iterate

Hello, I am trying to evaluate the performance of diff_iterate. This will affect our following decision on achieve "Copy local rbd image to remote ceph pool". Our current proposal is to first send a table which contains snapshot info and all the offset and length obtained from image.diff_iterate to remote site. Then remote site will request each extent from local site. Based on what we have tested, we know that diff_iterate is quite fast to return all extent changed. However, we would like to know more about how it works inside? Assume that a image has snap1(oldest), snap2 and snap3(latest). If we do a diff _iterate between snap3 to snap1, what does it do underneath? I suspect that for each snapshot, there is some sort of table that record all the extents for that snapshot. diffe_iterate is basically comparing 2 tables of different snapshots?? Am I correct? Please share more insights on this topic to help me get deeper understanding of diff_iterate and how a snapshot is created. Thanks, Jon Liu