FDP WAF Stress Result Summary:
작성일: fdp/reports/2026-06-15_poc_harness_evaluation_scenario.md
Run output: <output-root-or-run-dir>
Overall Result
<PASS / PARTIAL / INVALID 한 줄 요약. 예: Separated FDP replay completed successfully and WAF was available.>
| Metric | Value |
|---|---|
| Result status | <PASS/PARTIAL/INVALID> |
| Mode(s) | <no_fdp/mixed/separated> |
| Run ID | <run-id> |
| Device path | <device_path> |
| Block device path | <block_device_path> |
| Workers | <worker_count> |
| Warmup iterations | <warmup_iterations> |
| Measurement iterations | <measurement_iterations> |
| Total replay processes | <count> |
| Process failures | <count> |
| Replay record failures | <count> |
| Operation errors | <count> |
| Total wall time | <seconds> |
| Warmup wall time | <seconds or N/A> |
| Measurement wall time | <seconds> |
Precondition
원문 PoC 기준에서는 clean-state WAF artifact를 피하기 위해 preconditioning 여부와 범위를 명시한다.
| Metric | Value |
|---|---|
| Precondition status | <used/not used/unknown> |
| Precondition type | <window-local/full-span/none/unknown> |
| R0 output | <R0_precondition_* path or N/A> |
| Namespace discard | <yes/no/unknown> |
| Offset bytes | <bytes> |
| Size | <GiB or auto> |
| Sequential fill | <yes/no> |
| Random overwrite | <yes/no> |
| Random passes | <count> |
| Vendor media counter during R0 | <collected/not collected> |
WAF Result
| Mode | Host write bytes delta | Host write delta | Media write bytes delta | Media write delta | WAF | Write amplification over host writes | WAF status |
|---|---|---|---|---|---|---|---|
<mode> | <bytes> | <GiB> | <bytes> | <GiB> | <ratio> | <percent> | <available/unavailable: reason> |
Counter Collection
| Counter | Source | Status |
|---|---|---|
| Host writes | <nvme smart-log command> | <collected/not collected/fallback> |
| Media writes | <vendor media command> | <collected/not collected> |
| FDP stats | <nvme fdp stats or xnvme command> | <collected/not collected> |
| FDP RUH logs | <xnvme log-ruhu/fdp-ruhs> | <collected/not collected> |
Notes:
<counter caveat, e.g. xnvme unavailable; skipped FDP logs><SMART JSON/plain-text fallback note if any>
Measurement Workload Pressure
| Workload | Workers | Class | FDP Data RUHs | FDP Metadata RUHs | Logical Store Writes | Store records | Prefetch records |
|---|---|---|---|---|---|---|---|
llama8b_chat_chunk256 | <n> | <class> | <ids> | <ids> | <GiB> | <count> | <count> |
llama70b_longctx_chunk1024 | <n> | <class> | <ids> | <ids> | <GiB> | <count> | <count> |
rag_shared_prefix_chunk512 | <n> | <class> | <ids> | <ids> | <GiB> | <count> | <count> |
random_prompts_chunk128 | <n> | <class> | <ids> | <ids> | <GiB> | <count> | <count> |
metadata_heavy_small_objects | <n> | <class> | <ids> | <ids> | <GiB> | <count> | <count> |
| Total | <n> | <GiB> | <count> | <count> |
Latency
| Mode | write p50 ms | write p99 ms | read p50 ms | read p99 ms | all p99 ms |
|---|---|---|---|---|---|
<mode> | <ms> | <ms> | <ms or N/A> | <ms or N/A> | <ms> |
Log Observations
Replay completed <cleanly / with warnings / with errors>.
| Warning Category | Count | Files |
|---|---|---|
| Cleaning up in-flight store task | <count> | <count> |
| Failed batched memory allocation | <count> | <count> |
| Device metadata slot_bytes mismatch | <count> | <count> |
| FDP metadata payload rejected | <count> | <count> |
| io_uring fixed-buffer registration disabled | <count> | <count> |
| Tracebacks | <count> | <count> |
| ERROR logs | <count> | <count> |
| Exceptions | <count> | <count> |
Validity Against Scenario
| Criterion | Required by scenario | Observed | Status |
|---|---|---|---|
| Original goal: lifetime separation | separated RUH mapping separates hot/cold/metadata | <observed> | <PASS/FAIL/N/A> |
| Original goal: WAF available | host and media counters collected | <observed> | <PASS/FAIL> |
| Original goal: clean-state caveat addressed | R0 precondition status documented | <observed> | <PASS/FAIL/PARTIAL> |
| README scale | stress | <observed> | <PASS/FAIL> |
| README warmup | 2 | <observed> | <PASS/FAIL> |
| README measurement | 8 | <observed> | <PASS/FAIL> |
| Process failures | 0 | <observed> | <PASS/FAIL> |
| Replay record failures | 0 | <observed> | <PASS/FAIL> |
| Host write delta | > 0 | <observed> | <PASS/FAIL> |
| Media write delta | > 0 | <observed> | <PASS/FAIL> |
| p99 latency available | write p99 at minimum | <observed> | <PASS/FAIL/PARTIAL> |
Comparison
Use this section when comparing with another run, for example /home/ksy/waf-fdp-separated.
| Metric | Reference run | This run | Match |
|---|---|---|---|
| Mode | <value> | <value> | <yes/no> |
| Warmup iterations | <value> | <value> | <yes/no> |
| Measurement iterations | <value> | <value> | <yes/no> |
| Host write delta | <GiB> | <GiB> | <same/different> |
| Media write delta | <GiB> | <GiB> | <same/different> |
| WAF | <ratio> | <ratio> | <same/different> |
| WAF status | <value> | <value> | <yes/no> |
| Precondition | <value> | <value> | <yes/no/unknown> |
Conclusion
<One or two paragraphs. State what is proven, what is not proven, and the next required run.>
Recommended wording:
This run confirms that <mode/batch> completed without replay failures and that WAF was <available/unavailable>.
The result is <valid/partial/invalid> against the 2026-06-15 scenario because <reason>.
The next required step is <next run>.