{
  "slug": "deduplication-ratio",
  "term": "Deduplication ratio",
  "aliases": [
    "Data deduplication ratio",
    "Dedup ratio"
  ],
  "definition": "Deduplication ratio compares bytes required without duplicate sharing with bytes required after duplicate sharing for the same stated population and accounting scope.",
  "note": "Scope: This ratio isolates deduplication rather than crediting compression or unwritten thin-provisioned space. Declare which stored versions, matching domain and overhead are included. SNIA distinguishes deduplication-only reduction from other space-saving effects. [ratio-snia]\n\nPayload convention: For exact matching with one stored copy per distinct identity, divide all occurrence bytes by unique payload bytes. For equal-length chunks, the occurrence-count ratio gives the same result; otherwise calculate from byte totals rather than assuming count ratios are equivalent. References, indexes, replicas and allocation padding are outside this payload-only calculation; include them separately for an all-in footprint.\n\nRecipe cost: A reference per occurrence grows with logical content even when unique payload stays fixed. Recipe compression changes that cost. [ratio-recipes]\n\nSelected objects: A reported deduplication ratio may cover only deduplication-eligible objects. That selection is not automatically the compressible-data population used by a reported compression ratio. [ratio-vsan]\n\nStage boundary: Let L be all occurrence bytes, U unique bytes before compression, and C their compressed bytes. For positive denominators, (L / U) * (U / C) = L / C. This cancellation requires the same U, time and accounting boundary; separately reported dashboard ratios do not establish those conditions.\n\nLimits: A zero denominator does not provide a finite sizing ratio. The ratio alone establishes neither restore locality nor the additional bytes freed by deleting one retained version. Shared content can remain referenced by another version.",
  "related": [
    "data-reduction-ratio",
    "deduplication",
    "compression-ratio",
    "content-based-sampling"
  ],
  "citations": [
    {
      "id": "ratio-snia",
      "title": "SNIA Dictionary: compression ratio and data deduplication ratio, generated 17 November 2026",
      "url": "https://www.snia.org/sites/default/files/snia-dictionary-pdf/SNIADictionary.pdf",
      "accessed": "2026-11-17"
    },
    {
      "id": "ratio-vsan",
      "title": "Broadcom KB 315553: vSAN ESA 9.1 space-efficiency counter populations",
      "url": "https://knowledge.broadcom.com/external/article/315553",
      "accessed": "2026-11-17"
    },
    {
      "id": "ratio-recipes",
      "title": "Meister, Brinkmann and Suss, File Recipe Compression in Data Deduplication Systems, FAST 2013, sections 1-3",
      "url": "https://www.usenix.org/system/files/conference/fast13/fast13-final54.pdf",
      "accessed": "2026-11-17"
    }
  ],
  "status": "unknown",
  "advantages": [],
  "limitations": [],
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "updated": "2026-11-22",
  "provenance": [
    "The L/U and U/C example is dimensional accounting for a stated deduplication-then-compression model, not a claim about every product's processing order.",
    "No measured array result, deduplication benchmark or capacity guarantee is implied.",
    "The selected-population example follows Broadcom's vSAN ESA 9.1 counter definitions, not a universal convention. [ratio-vsan]"
  ]
}
