{
  "slug": "unified-buffer-management",
  "term": "Unified Buffer Management (UBM)",
  "aliases": [
    "UBM buffer management"
  ],
  "definition": "Unified Buffer Management classifies file-block accesses and allocates cache space among sequential, looping and other-reference partitions.",
  "note": "Detection, replacement and allocation are separate modules. Per-file consecutive blocks identify sequences; repeated sequences establish loops. Sequential blocks use MRU, loops use period-based replacement with MRU ties, and the paper uses LRU for other references. [ubm]\n\nFor a loop of l blocks with period p, the model gives Hit(n)=min(l,n)/p. Its marginal gain is 1/p through n=l and zero beyond saturation. Time here counts all block references, not seconds. Comparing partitions requires the same observation clock. [ubm]\n\nIn a constructed four-block loop with period 20, two buffers yield 2/20=0.1 hits per global reference, although 2/4=0.5 of that loop's accesses hit in the model. Doubling the period to 40 halves the former rate without changing the latter fraction. A fifth buffer adds no hits after all four loop blocks fit. [example]\n\nA newly detected sequence initially lacks loop evidence. Under cache pressure, its first repeated access can miss before the detector recognizes looping; cold-cache spare space can conceal this cost. Detection therefore does not promise a first-reuse hit. [ubm]",
  "status": "research",
  "advantages": [],
  "limitations": [
    {
      "text": "Estimated reference-count gains do not price unequal I/O latency, dirty writeback or prefetch costs.",
      "citationIds": [
        "ubm"
      ]
    }
  ],
  "related": [
    "program-counter-based-classification",
    "locality-of-reference"
  ],
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "updated": "2026-11-21",
  "citations": [
    {
      "id": "ubm",
      "title": "Kim et al. A Low-Overhead High-Performance Unified Buffer Management Scheme that Exploits Sequential and Looping References. OSDI 2000, sections 3-5.",
      "url": "https://www.usenix.org/legacy/event/osdi00/full_papers/kim/kim_html/index.html",
      "accessed": "2026-11-21"
    },
    {
      "id": "example",
      "title": "Hesela. Loop-allocation arithmetic and explicit global-reference denominators.",
      "url": "https://hesela.dev/research/ubm/examples.json",
      "accessed": "2026-11-21"
    },
    {
      "id": "code",
      "title": "Hesela. Bounded loop expectation calculation and deterministic example generator.",
      "url": "https://hesela.dev/research/ubm/model.mjs",
      "accessed": "2026-11-21"
    },
    {
      "id": "tests",
      "title": "Hesela. Loop expectation boundary tests and exhaustive event-count oracle.",
      "url": "https://hesela.dev/research/ubm/model.test.mjs",
      "accessed": "2026-11-21"
    },
    {
      "id": "guide",
      "title": "Hesela. Reproduction instructions and model scope.",
      "url": "https://hesela.dev/research/ubm/README.txt",
      "accessed": "2026-11-21"
    }
  ],
  "provenance": [
    "OSDI 2000 combines trace-driven simulation and a FreeBSD prototype. Its workload-specific improvements are not forecasts for modern systems. [ubm]",
    "Synthetic arithmetic, not a native UBM implementation, measured trace or reproduction of the paper's reported performance. [example]"
  ]
}
