{
  "aliases": [
    "Residual service curve",
    "Blind-multiplexing service bound"
  ],
  "advantages": [],
  "limitations": [],
  "status": "shipping",
  "supersedes": [],
  "supersededBy": [],
  "firstSeen": null,
  "lastSeen": null,
  "updated": "2026-11-20",
  "slug": "leftover-service-curve",
  "term": "Leftover service curve",
  "definition": "A leftover service curve lower-bounds the service available to a selected flow after accounting for competing traffic.",
  "note": "Blind multiplexing: strict aggregate beta and cross-traffic envelope alpha2 yield the ordinary residual curve max(0,beta-alpha2) when that curve is nondecreasing. This accounts for bursts, not just average utilization. [netcal-book]\n\nFor strict beta(u)=R*max(0,u-T), alpha2(u)=b2+r2*u for u>0 and r2<R, the residual rate is R-r2 and latency is (R*T+b2)/(R-r2). A target envelope with rate at most R-r2 then admits finite backlog and virtual-delay bounds. [netcal-book]\n\nThe general arbitrary-multiplexing construction uses the nondecreasing upper closure of beta-alpha2. [dnc2026] A residual ordinary curve is not automatically strict; serial composition and subtraction must retain their respective hypotheses. [negative-service]",
  "related": [
    "strict-service-curve",
    "rate-latency-service-curve",
    "arrival-curve",
    "token-bucket",
    "tail-latency"
  ],
  "citations": [
    {
      "id": "netcal-book",
      "title": "Le Boudec and Thiran. Network Calculus, 2022 online edition, Definition 1.3.2 and Theorem 6.2.1.",
      "url": "https://leboudec.github.io/netcal/latex/netCalBook.pdf",
      "accessed": "2026-11-20"
    },
    {
      "id": "negative-service",
      "title": "Hamscher, Constantin and Schmitt. Extending Network Calculus To Deal With Partially Negative And Decreasing Service Curves (2024), sections I-II.",
      "url": "https://arxiv.org/pdf/2403.18042v1",
      "accessed": "2026-11-20"
    },
    {
      "id": "dnc2026",
      "title": "Pesotsky, Hermsen and Bondorf. Automated and Precise Deterministic NetCal Calculations from Models to Bounds. ECRTS 2026, Theorem 10.",
      "url": "https://drops.dagstuhl.de/doi/10.4230/LIPIcs.ECRTS.2026.6",
      "accessed": "2026-11-20"
    },
    {
      "id": "hesela",
      "title": "Hesela: Cross traffic consumes burst budget as well as rate",
      "url": "https://hesela.com/analyses/storage-tail-slowdowns-raid-fanout/#leftover-service",
      "accessed": "2026-11-20"
    }
  ],
  "provenance": [
    "Published theory and a linked exact hypothetical calculation, not a storage-device benchmark. Individual delay needs a within-flow ordering guarantee."
  ]
}
