Semantic trace playback
Semantic trace playback evaluates candidate filesystem policies by transforming recorded I/O and issuing the resulting requests to storage.
Timing-sensitive changes need more than a block trace. STP also records buffer-dirtying activity and application fsync calls. These distinguish synchronization requests from timer- or threshold-triggered writes. [1]
A faster replay can justify investigating a policy; it does not establish that an implementation preserves crash recovery or application correctness.
Relations
Limitations
- Unmodeled virtual-memory interactions and radical filesystem redesigns can invalidate the playback approximation. [1]
Citations
Sources & caveats
- Primary-source synthesis of section 2.2; the implementation/correctness distinction is an editorial inference, not a reproduced benchmark.