Linux storage tiering: Difference between revisions

From Roy's somewhat wise thoughts
Jump to navigation Jump to search
 
Line 1: Line 1:
= Tiering =
= Tiering =


Tiering eller lagdelt lagring er et begrep som brukes om datalagring der man i stedet for å velge å lagre på teip, treige eller raske disker, SSD-er, NVME-er eller en rein RAM-disk, har en kombinasjon av flere av disse.
Storage tiering is like caching, except that with caching, the cached data is stored both on the slower medium (typically rotating drives) and in cache (like SSDs or NVMEs). With tiering, the data is migrated between the tiers according to usage statistics. This means tiering is a bit like persistent cache, but also frees up the space used by the data in the cache. In many usecases, this won't be necessary, since the cache size usually is only a small fraction of the slower storage. Some large-scale storage solution also use magnetic tape as the bottom tier, spooling off data not used for some time out tape with a large tape robot. This saves space and power and ensures the data is available, albeit perhaps with a wee bit of delay for that on tape.


== autotier ==
== autotier ==
Line 15: Line 15:


== ZFS ==
== ZFS ==
ZFS støtter ikke tiering direkte, men har fra rundt 2020(?) caching som overlever en omstart.
ZFS does not support tiering, but has had persistent L2ARC (read cache) since around 2020.

Latest revision as of 15:24, 29 April 2022

Tiering

Storage tiering is like caching, except that with caching, the cached data is stored both on the slower medium (typically rotating drives) and in cache (like SSDs or NVMEs). With tiering, the data is migrated between the tiers according to usage statistics. This means tiering is a bit like persistent cache, but also frees up the space used by the data in the cache. In many usecases, this won't be necessary, since the cache size usually is only a small fraction of the slower storage. Some large-scale storage solution also use magnetic tape as the bottom tier, spooling off data not used for some time out tape with a large tape robot. This saves space and power and ensures the data is available, albeit perhaps with a wee bit of delay for that on tape.

autotier

btrfs

bcachefs

  • bcache, a caching engine for the Linux block level, came into mainline Linux Kernel in 2013 and has been regarded as stable for quite some time now (2022-04-29).
  • bcachefs is a filesystem built on the same codebase as bcache, but
  • Documentation

ZFS

ZFS does not support tiering, but has had persistent L2ARC (read cache) since around 2020.