Btrfs and called a new filesystem that Oracle is developing under the GPL. Still in very early stage of development and is very functional, which is supposed to change as development goes ahead. This is a list of features that offer when considering stable and goes into production: very large file storage (2 64 maximum file size)
efficiency in storing small files
storage efficiency in the indexed directories
Dynamic allocation of i-nodes Snapshots modifiable (!)
subvolumens (!)
Mirroring and striping at the object level (!)
Checksums on data and metadata tight integration with the mapper devices to support multiple devices
- hot Checking filesystem
- very fast filesystem Verification cold
- efficient incremental backups (!) and FS mirroring
- It may be a kind of mix to bring ASM to a Linux filesystem, so it makes me think of 'subvolumens' and mirroring and striping at the object level (this makes me drool!). If I am not mistaken, the object-level striping + the + devices support multiple dynamic inode allocation would allow the system itself was able to detect objects that are accessed constantly moving so that the burden of I / O in various devices, hot and transparently.
- The efficient incremental backups also has a point, because if the filesystem is able to say a priori that is what has changed - and if that means that mark ("bitmaps?) Only blocks that have changed, no the entire file - we will have a dramatic improvement in backup times. If the above is true, I get the feeling that we have a filesystem with touches of what we already have in Oracle database: local management / automatic space MultiDevice subvolumens + + strip of objects (concepts of tablespaces and ASM), backups 'Retailers' (concepts of RMAN?)
- Sources: Project Home
-
Btrfs - I will ask to see if I get more info on these features, and'll tell (on ask around with * means Btr)
0 comments:
Post a Comment