Oversimplified but I am trying to be concise.
In conventional magnetic recording (CMR), there is a one-to-one sequential relationship between the logical block numbers (LBN) the operating system understands, and the actual physical location of the blocks in the hard drive - data blocks on the hard drive are never moved internally in CMR. Deletes are handled entirely by the operating system by flagging the deleted logical blocks as deleted - the drive never needs to know about the deleted blocks and the old data in the block is not erased on the drive. The OS will just overwrite deleted flagged logical blocks as needed, which overwrites the old data. This is why deleted data can be often recovered on CMR drives.
The Windows OS works the same way with SSD and SMR, except these drives do move data around internally which causes some issues and necessitates support for deletes in the TRIM command, which is sent to the drive by the host (dynamically with NTFS).
SSD drives and device based SMR HDDs need to move the data around internally for space management and performance purposes. They maintain an internal mapping between the LBN known by the OS and the actual physical block the data is actually stored in, so there is an additional level of indirection compared to CMR. But now the drives need to know about deletes so they can manage space internally, because the OS now cannot. This is the reason for the TRIM command. When you delete blocks that are on an SSD for example, the TRIM command sends a list of the logical blocks that are deleted to the SSD. The SSD then maps the LBN to the physical block and flags it as deleted so it can reuse that space.
Shingled magnetic recording was first introduced on magnetic tape media as far as I am aware (I was responsible for the management of hundreds of Petabytes of shingled tape). Writing data onto tracks on a magnetic medium has a physical limitation on the minimum width of a track due to the magnetic field generated by the head to magnetize the media. Reading a track can be narrower because there is no magnetic field being generated by the head. Shingled recording takes advantage of this condition by overlapping tracks when writing. Each consecutive track is overlapped with the previous track. This results in readable adjacent tracks of approximately half the width the write head can write, and they can be read fine. The trouble comes when you want to overwrite a block on track. If you write to the 1/2 width track with the write head, it will also overwrite an adjacent tracks data. too This is what makes them very different from CMR.
Device based SMR divides the disk into zones which they treat as a unit for writing. They also reserve disk space for a cache that gets directly written to from the hoist, then moved to a CMR zone when the drive is idle. Anyhow, there is no one to one mapping to the data like SSD, but it is quite different. I don't know much about the internals of this since I have never really used SMR drives and never will. They were created when there was a technology gap limiting the capacity of hard drives. But then helium filled drives that permitted platters to be closer together and hence more platters soon came into being. You can get very high capacity CMR drives now. The only thing SMR has going is they are slightly cheaper. It seems in the consumer space, the niche for SMR is in the low-end smaller capacity drives, just for cost savings.
Cheers
Apple 13" Macbook Pro 2020 (m1)
