We have some volumes with Storage Replica running between them. When writing data to the source volume it looks like the read activity from the source log is ~3 times larger than the actual data throughput. The effect becomes larger when individual changes to the fileshare are smaller.
Is there a minimum block size that SR uses regardless of the change? We can see this behaviour across iSCSI drives in a stretch cluster as well as on local disks in server-to-server replication.