“Writes to a region that has already been copied will be mirrored to the source and destination – as the DM has already passed this area and any new updates must be reflected by the IO Mirror. Writes to the region currently being copied (in between the two offsets) will be deferred and placed into a queue. Once the DM IO completes we
enqueue those writes and unlock the region by updating the offsets. As part of the updating operation we wait for inflight writes to complete. The final region is not mirrored and all writes are issued to the source only”
I would like to know, for not mirrored scene, whether the following optimization can be carried out？
write request is not migrated and the size of the new write request data block is equal to the size of the migration data block，directly writing the new write request data block
into the source and the destination，and update has been migrated. Because it can reduce the operation of reading the metadata at one time, reduce the storage bandwidth occupancy of the source storage, Reduced impact on business IO performance。
And If the size of the data block is not same to the migrate data, all writes are issued to the source only.