Product SiteDocumentation Site

Chapter 6. Allocators

6.1. Allocation Policy
6.2. Allocation Policy - Directories
6.3. Allocation Policy - Files
6.4. Inode Numbers
6.5. Inode Number Format
6.6. Inode Number Size
6.7. 32bit and 64bit Inodes
6.8. 32bit and 64bit Inodes
6.9. 32 bit Inodes on >1TB Filesystems
6.10. Rotor Step
6.11. Realtime Allocator
6.12. Realtime Allocator Limitations
6.13. Traditional allocator impact on some workloads
6.14. Traditional Stream Allocation
6.15. RAID performance with interleaved streams
6.16. Filestreams Allocator
6.17. RAID performance with filestreams
6.18. Fragmentation
6.19. Fragmentation Example
6.20. xfs_fsr

6.1. Allocation Policy

The default allocation behaviour of XFS for new files is to place them in the same allocation group as their parent directory.
Since files and their parent directories are often accessed in close succession, this minimises costly disk seeks.
The allocator will also attempt to place newly created directories in different allocation groups.
Combined, these policies help group directories of files together on disk, even if they're being written to concurrently.
Allocation policies in XFS can change due to
  • 32 bit inode numbers are used on large file systems
  • Some mount options