Soft Updates: A Solution to the Metadata Update Problem in File Systems

Reading notes

Metadata: directories, inodes, and free block maps and so on
Metadata updates: file creation, block allocation and so on
Soft updates: an implementation technique for low-cost sequencing of fine-grained updates to write-back cache blocks. So, a file system can safely use delayed writes for almost all file operations.
(unpredictable system crashes -> loss of all information in volatile main memory)
On-disk image of the file system must have
- no dangling pointers to uninitialized space,
- no ambiguous resource ownership caused by multiple pointers
- no live resources to which there are no pointers.
Ordering constraints:
- (1) Never point to a structure before it has been initialized (e.g., an inode must be initialized before a directory entry references it).
- (2) Never reuse a resource before nullifying all previous pointers to it (e.g., an inode’s pointer to a data block must be nullified before that disk block may be reallocated for a new inode).
- (3) Never reset the last pointer to a live resource before a new pointer has been set (e.g., when renaming a file, do not remove the old name for an inode until after the new name has been written).

immediate stability and consistency of all metadata updates with no restrictions on on-disk data organization, no performance overhead, and no special hard- ware support
Performance-related characteristics of an ideal solution:
1. Applications should never wait for disk writes unless they explicitly choose to do so for application-specific purposes.
2. The system should propagate modified metadata to disk using the minimum possible number of disk writes, given the allowed window of vulnerability. Specifically, this requires aggressive write-back caching of metadata structures to absorb and coalesce writes.
3. The solution should minimize the amount of main memory needed to cache dirty metadata and related auxiliary information. This will maximize the availability of memory for other purposes.
4. The cache write-back code and the disk request scheduler should not be constrained in choosing what blocks to write to disk when, beyond the minimal restrictions necessary to guarantee consistency. This flexibil- ity is important for scheduling algorithms that reduce mechanical positioning delays [Denning 1967; Worthington et al. 19

When creating a new file, the newly initialized inode must be written to disk before the new directory entry.
When removing a file, the reset directory entry must be written before the reinitialized inode.