{
  "slug": "storage-tail-slowdowns-raid-fanout",
  "title": "Slow RAID hours are not request tail latency",
  "kind": "dataset",
  "summary": "A slow-drive hour, a slow-RAID hour and a request percentile count different events. This source review separates them and shows why fan-out estimates need explicit dependence and workload assumptions.",
  "license": "No standalone reuse license has been verified for this mixed-source derived CSV. It contains attributed figures and labeled arithmetic, not the papers or raw fleet traces. Each source retains its own rights.",
  "licenseUrl": "https://hesela.dev/datasets/storage-tail-slowdowns-raid-fanout/#rights-and-reuse",
  "sourceUrl": "https://www.usenix.org/system/files/conference/fast16/fast16-papers-hao.pdf",
  "accessed": "2026-10-15",
  "rowCount": 24,
  "columns": [
    {
      "name": "measure",
      "type": "string",
      "description": "The quantity as the source names it, or our derived quantity."
    },
    {
      "name": "value_low",
      "type": "number",
      "description": "Legacy lower endpoint: some rows pair HDD and SSD values rather than a statistical interval; empty for point values."
    },
    {
      "name": "value_high",
      "type": "number",
      "description": "Point value or paired upper endpoint; never an implied uncertainty bound."
    },
    {
      "name": "unit",
      "type": "string",
      "description": "Unit of the value columns: drives, drive hours, percent, milliseconds, times, reports."
    },
    {
      "name": "scope",
      "type": "string",
      "description": "Population and window the value applies to; 'Derived by us' marks our arithmetic."
    },
    {
      "name": "note",
      "type": "string",
      "description": "What the value is not, or the source wording behind it."
    },
    {
      "name": "citation_id",
      "type": "string",
      "description": "Id of the opened source in the citations list."
    }
  ],
  "downloadPath": "/data/datasets/storage-tail-slowdowns-raid-fanout.csv",
  "humanPage": "https://hesela.com/analyses/storage-tail-slowdowns-raid-fanout/",
  "limits": [
    "Hourly summaries cannot recover request percentiles, simultaneous sub-hour stalls, or an empirical confidence interval that accounts for temporal and group dependence. A conditional mean above one establishes multiple slow classifications within some group-hours, not statistical dependence between drives. Historical fleets, operator reports and selected benchmark traces must not be pooled. No current SSD failure rate, universal RAID slowdown, measured mitigation benefit or vendor ranking follows.",
    "The retained value_low/value_high columns also hold HDD/SSD category endpoints, not confidence intervals. Read the scope before comparing values.",
    "The 24 table rows are an editorial extract, not the empirical sample size. Source publication dates are not collection dates."
  ],
  "citations": [
    {
      "id": "hao-fast16-pdf",
      "title": "Hao et al., The Tail at Store, USENIX FAST 2016, pp. 263-276; Table 1, sections 3, 4.1.2, 5.2 and 6",
      "url": "https://www.usenix.org/system/files/conference/fast16/fast16-papers-hao.pdf",
      "accessed": "2026-10-15"
    },
    {
      "id": "dean-barroso-cacm13-pdf",
      "title": "Dean and Barroso, The Tail at Scale, CACM 56(2), 2013, pp. 74-80; Tables 1-2; DOI 10.1145/2408776.2408794",
      "url": "https://www.barroso.org/publications/TheTailAtScale.pdf",
      "accessed": "2026-10-15"
    },
    {
      "id": "gunawi-fast18-pdf",
      "title": "Gunawi et al., Fail-Slow at Scale, USENIX FAST 2018; report collection (section 2) and detection time (section 3.5)",
      "url": "https://www.usenix.org/system/files/conference/fast18/fast18-gunawi.pdf",
      "accessed": "2026-10-15"
    },
    {
      "id": "lu-fast23-pdf",
      "title": "Lu et al., Perseus, USENIX FAST 2023, pp. 49-64; abstract, sections 4.4-4.5 and 5.1",
      "url": "https://www.usenix.org/system/files/fast23-lu.pdf",
      "accessed": "2026-10-15"
    }
  ],
  "updated": "2026-10-15",
  "method": "Source audit of four full papers, with locators in the table notes and accompanying dataset. Published rounded figures are retained separately from Hesela's reproducible arithmetic. Table 1 of Hao et al. supplies the integer exposure and slow-hour counts; no per-request traces were obtained. Synthetic examples use explicit values, exact fractions and nearest-rank percentiles. They are counterexamples, not a replication of a production fleet.",
  "rightsStatus": "not-verified",
  "rightsScope": "derived-csv",
  "rightsReviewed": "2026-10-15",
  "rightsEvidence": [
    {
      "title": "Hao et al., The Tail at Store, USENIX FAST 2016, pp. 263-276; Table 1, sections 3, 4.1.2, 5.2 and 6",
      "url": "https://www.usenix.org/system/files/conference/fast16/fast16-papers-hao.pdf",
      "accessed": "2026-10-15",
      "note": "Reviewed for factual provenance; the publication is not a verified standalone reuse license for this mixed-source CSV."
    },
    {
      "title": "Dean and Barroso, The Tail at Scale, CACM 56(2), 2013, pp. 74-80; Tables 1-2; DOI 10.1145/2408776.2408794",
      "url": "https://www.barroso.org/publications/TheTailAtScale.pdf",
      "accessed": "2026-10-15",
      "note": "Reviewed for factual provenance; the publication is not a verified standalone reuse license for this mixed-source CSV."
    },
    {
      "title": "Gunawi et al., Fail-Slow at Scale, USENIX FAST 2018; report collection (section 2) and detection time (section 3.5)",
      "url": "https://www.usenix.org/system/files/conference/fast18/fast18-gunawi.pdf",
      "accessed": "2026-10-15",
      "note": "Reviewed for factual provenance; the publication is not a verified standalone reuse license for this mixed-source CSV."
    },
    {
      "title": "Lu et al., Perseus, USENIX FAST 2023, pp. 49-64; abstract, sections 4.4-4.5 and 5.1",
      "url": "https://www.usenix.org/system/files/fast23-lu.pdf",
      "accessed": "2026-10-15",
      "note": "Reviewed for factual provenance; the publication is not a verified standalone reuse license for this mixed-source CSV."
    }
  ],
  "reproduction": {
    "code": "/research/tail-sampling/model.py",
    "tests": "/research/tail-sampling/test_model.py",
    "json": "/research/tail-sampling/results.json",
    "command": "python3 model.py --output results.json",
    "runtime": "Python 3.12 or newer; standard library only",
    "kind": "Exact arithmetic on printed counts plus synthetic counterexamples; no fleet trace reanalysis."
  }
}