{
  "slug": "work-conserving-scheduler",
  "term": "Work-conserving scheduler",
  "aliases": [
    "work-conserving scheduling"
  ],
  "definition": "A work-conserving scheduler does not deliberately leave its service resource idle while requests are waiting to be served.",
  "note": "Anticipatory scheduling can deliberately wait for a synchronous process's next demand request even when other requests are pending. Such non-work-conserving behavior can reduce seeks or protect proportional service; keeping a disk busy is not the same objective as completing useful work efficiently. [iyer2001]\n\nState the scheduling boundary. Linux blk-mq separates software staging from hardware dispatch, and resource shortages can delay submission. A gap in host dispatch alone does not establish that all device resources are idle or that a policy deliberately withheld work. This is a measurement implication of the layered path. [blk-mq]",
  "status": "research",
  "advantages": [],
  "limitations": [
    {
      "text": "Work conservation alone specifies neither service order nor fairness between requesters.",
      "citationIds": [
        "iyer2001"
      ]
    }
  ],
  "related": [
    "io-scheduler",
    "anticipatory-scheduling",
    "bfq"
  ],
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "updated": "2026-11-08",
  "citations": [
    {
      "id": "iyer2001",
      "title": "Iyer and Druschel. Anticipatory scheduling. SOSP 2001, sections 2 and 3.2.",
      "url": "https://www.sosp.org/2001/papers/iyer.pdf",
      "accessed": "2026-11-08"
    },
    {
      "id": "blk-mq",
      "title": "Linux 6.16: blk-mq, operation and hardware dispatch queues",
      "url": "https://docs.kernel.org/6.16/block/blk-mq.html",
      "accessed": "2026-11-08"
    },
    {
      "id": "hesela",
      "title": "Hesela: Sharing an SSD: who pays for the slowdown?",
      "url": "https://hesela.com/analyses/ssd-fairness-shared-queues/",
      "accessed": "2026-11-08"
    }
  ],
  "provenance": [
    "Primary-source synthesis, not a new scheduler implementation or benchmark.",
    "The historical disk results do not establish performance gains for contemporary SSDs."
  ]
}