Device-managed SMR (DM-SMR)
Device-managed SMR presents a conventional logical-block interface while drive firmware manages data placement and updates on shingled media.
Drive-managed SMR accepts ordinary block writes without exposing its band layout as host-managed zones. Firmware performs placement through a shingle translation layer; accepting random writes does not promise CMR-like latency. [6]
Linux v6.8 also treats host-aware disks as conventional. The queue/zoned value none therefore identifies neither CMR nor DM-SMR uniquely. The separate SCSI zoned_cap attribute reflects advertised capability when available. [8] [9]
Persistent media staging differs from a volatile write buffer. Cleaning can compete with pending requests. [10] A slowdown alone does not establish a command failure. [7]
Relations
- related
- Shingled magnetic recording (SMR), Host-managed SMR (HM-SMR), Nearline HDD, SMR persistent cache, SMR garbage collection, Shingled band, Host-aware SMR, Shingle translation layer
Advantages
- Works with legacy software unaware of zoned semantics. [6]
Limitations
- Sustained random writes can trigger internal garbage collection and latency spikes. [5]
Citations
- Aghayev and Desnoyers: Skylight, FAST 2015, sections 4.5-4.6
- Zoned Storage: Shingled Magnetic Recording Hard Disks
- Hesela: SMR stalls, cache limits and the errors SMART can miss
- Linux v6.8: sd_read_block_characteristics and zoned_cap_show
- Linux v6.8: queue_zoned_show
- Aghayev et al.: Evolving Ext4 for Shingled Disks, FAST 2017, section 4.3
Further reading
- Shingled magnetic recording
- Zoned Storage — SMR disk models (drive-managed vs zoned)
- Zoned Storage — SMR introduction (drive-managed model)
- Western Digital — WD Red SMR vs CMR NAS drive information
Sources & caveats
- Linux claims are pinned to v6.8 and do not make every drive or bridge report a useful zoned_cap value.
- The linked analysis separates published device observations from synthetic resource arithmetic.