NotesUnderstanding Flat File Systems
A fat file system consists of an array of index nodes (or “i-nodes” or “inode”)
There is one i-node per file, which contains
File Allocation Table (FAT)
Indexed Allocation
Free Space Management
Bit vectors:
- A bit for each block, 1 for available, 0 otherwise
Linked Lists:
- Link all free blocks in one list
Sparse Files
This is conceptual, but a file can have empty “holes”. Supposedly better for debugging or something. Useless, but supported in file systems.
You can’t use this with sequential or linked allocations.
File Blocks
Not exactly a disk or SSD sector.
Tradeoffs are similar to pages:
- If we have larger blocks, we can locate them quicker
- If we have larger blocks, we don’t have to do a lot of disk accesses to get an inode
Optimization for SSDs
Exploit the low latency
- No need to worry about disk seeks or rotational latency
- Reduce the number of writes or amount of data to be written (e.g., use compression)
Use the TRIM command
- SSD controllers rework the allocation to reduce wear leveling and do garbage collection to reduce the write inflation factor
File Extent
Allocate files in extents (what is an extent??)
An extent is a group of consecutive block numbers
Metadata: a start and number of blocks
More efficient for access and lower overhead (used by EXT4, APFS, etc.)