Persistence ordering and repeatable recovery

This required 22–32-hour package replaces the former three-hour storage simulator exercise: lessons and predictions 6–8, implementation and crash experiments 10–16, assessment and delayed repair 6–8. Estimates await learner pilots. Existing filesystem layouts, descriptors, devices, page cache, readiness and asynchronous I/O practice remains required. See the reconciled semester workload.

Prerequisite bridge

Bring Semester 04's distinction between accepted writes and durable effects, Module 2's separation of cached and authoritative state, and Module 3's precise operation boundaries. Draw two copies of a value, label volatile/durable, and explain assignment versus increment. Use 1–2 hours inside lessons to repair gaps.

Learning sequence

1. Durability and acknowledgement.
2. Record boundaries and replay.
3. Crash matrices and storage assumptions.
4. Prediction practice, runnable lab, FS1–FS6 and calibration.

Selective primary resources

Use OSTEP Crash Consistency for the ordering problem and journaling context. Consult Linux fsync(2) for the real API boundary, including errors and the separate containing-directory persistence obligation. Sources checked on 2026-10-23. The model does not call fsync or test a device. Original OSTEP/Operating System Concepts selections remain the backbone; no purchase or book conversion is required.