File systems and journaling

In short: A file system organises how files and folders are stored on a storage device. Journaling file systems (ext4, NTFS) log changes so they stay consistent after a crash.

In more detail: The disk is divided into blocks. A structure (inode or MFT entry) describes each file: size, permissions, location of blocks. Directories are files that map names to these entries. In journaling a change is noted in the journal first and then carried out.

In Depth

• Inode: Metadata of a file, without its name. Names live in the directory (hence hard links).
• Journal: Metadata only or data as well. Protects against half-finished changes.
• Copy-on-write file systems (ZFS, Btrfs): Never write in place, snapshots almost free, checksums detect silent data errors.
• Fragmentation: Files split into pieces; uncritical on SSDs.
• Encryption and quotas: At file system or block level.
• FAT32/exFAT: Simple, compatible, for USB sticks; no journal, FAT32 max. 4 GB per file.
• fsync: Forces writing to the device, needed for durability.

See also: Write-ahead log and recovery, Virtual memory and paging