Skip to main content

Advanced FEMU study path

Each study produces a reproducible artifact: hypothesis, frozen configuration, workload, raw output, checker, explanation, and a reset procedure. These are study assignments, not claims of freshly qualified guest labs.

ModuleInvestigationDeliverable
A0 FoundationsTrace NVMe command, page, block, mapping and eraseAnnotated request path and unit conversions
A1 MethodologySeparate occupancy, warm-up and measurementControlled comparison with repeated trials
A2 MappingCompare page mapping and DFTL translation trafficCache-size sweep including mapping writes
A3 PlacementVary lifetime separation, groups and handlesFDP configuration and interval byte accounting
A4 Zoned storageExplore active/open limits and write cachesState-machine tests and recovery trace
A5 KV storageCompare packing and index choicesKey/value correctness plus space/media accounting
A6 CSDCompare a real algorithm with a CPU referenceProgram hash, result proof and clock scope
A7 CXLSeparate CPU-cache and device-cache effectsRegion configuration, access trace and counters
A8 ReliabilityInject one supported fault with known ground truthExpected error and recovery invariant
A9 InstrumentationTrace one request across queues and timingSource-linked timeline without double counting
A10 ExtensionAdd a small tested interfaceRed/green patch and invalid-input regression
A11 ArtifactExchange a complete experiment with another learnerIndependent reproduction report
A12 Full stackRun one application with fixed input and configurationApplication correctness and device-level evidence
A13 CalibrationCompare a documented hardware behaviorCalibration envelope and held-out validation

Use the manual to choose supported features at your source pin. Do not compare incompatible interfaces as if they were the same experiment. DFTL under FDP and non-FDP have different accounting/model details at this edition's pin; state which path you exercised.