{
  "slug": "gl-cache",
  "term": "GL-Cache",
  "aliases": [
    "Group-level learned cache"
  ],
  "definition": "GL-Cache is a learned cache design that predicts the usefulness of object groups and combines group selection with merge-based eviction.",
  "note": "Objects are grouped by write time; a learned model ranks groups, while a lightweight age-and-size rule selects survivors during merging. Group-level learning therefore does not remove object-level decisions. [paper]\n\nThe utility target sums 1/(object size times next-request delay). Future requests supply training labels; eviction uses predictions, not advance knowledge. Unrequested objects can leave online labels below their complete-future values. Ghost entries extend observation after eviction. [paper]\n\nOnly the first merge group is selected by utility; the paper chooses companions close in write time. Selecting the globally lowest-ranked groups is a different policy. [paper]\n\nThe pinned C micro-implementation calculates training contributions using request-counter distance plus one, scaled by 1,000,000 and divided by object size. Its feature preparation separately uses real-time group age. These clocks and the label scale must not be silently interchanged when reproducing the algorithm. [training]\n\nThe author repository separates a metadata-only micro-implementation from a Rust prototype based on Pelikan Segcache. They are different evaluation surfaces, not interchangeable measurements of a deployed service. [repository]",
  "status": "research",
  "advantages": [],
  "limitations": [
    {
      "text": "The reported 228-fold throughput comparison with LRB is for the micro-implementation at small cache sizes. It excludes networking and value storage, so it is not an end-to-end service speedup.",
      "citationIds": [
        "paper"
      ]
    }
  ],
  "related": [
    "least-hit-density",
    "cacheus",
    "ghost-cache-entry",
    "segcache"
  ],
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "updated": "2026-11-19",
  "citations": [
    {
      "id": "paper",
      "title": "Yang, Mao, Yue and Rashmi. GL-Cache: Group-level learning for efficient and high-performance caching. FAST 2023, pp. 115-133, sections 3.3-3.6, 4.1 and 4.4.",
      "url": "https://www.usenix.org/system/files/fast23-yang.pdf",
      "accessed": "2026-11-19"
    },
    {
      "id": "training",
      "title": "GL-Cache author micro-implementation, dataPrep.c: prepare_one_row and update_train_y; commit fbb824091209f390f7c921aadee1fbf0d35209c2.",
      "url": "https://github.com/Thesys-lab/fast23-GLCache/blob/fbb824091209f390f7c921aadee1fbf0d35209c2/micro-implementation/libCacheSim/cache/eviction/GLCache/dataPrep.c",
      "accessed": "2026-11-19"
    },
    {
      "id": "repository",
      "title": "GL-Cache author repository: micro-implementation versus prototype; commit fbb824091209f390f7c921aadee1fbf0d35209c2.",
      "url": "https://github.com/Thesys-lab/fast23-GLCache/blob/fbb824091209f390f7c921aadee1fbf0d35209c2/README.md",
      "accessed": "2026-11-19"
    }
  ],
  "provenance": [
    "Primary-paper and pinned-source review. No trace replay, native benchmark, model training or deployment-performance replication was performed.",
    "The record distinguishes the paper's design from the released micro-implementation; code-specific clocks are not attributed to every implementation."
  ]
}
