(2.7), p. 7 — Q(t) = I(t) + W₀(t) + M₂(t) + A_G(t) + Σ (G(t − w̃ᵢ) − G(t)) + Σ (G(t − τᵢ − wᵢ) − G(t − τᵢ))
ProvedReedGGN.SystemEq.equation_2_7halfin-whittmany-server-queuep2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1queueing
Let and be the service-time and residual service-time distributions, with tails and , and fix a sample path of the queue. With the number in system (2.2) and , , , as in (2.4)–(2.6) and p. 7, for every ,
The paper obtains it from (2.2) by centring each indicator at its (conditional) mean, which gives (2.3), and then adding and subtracting and . It separates the number in system into a mean term for the initial customers, two fluctuation terms, the infinite-server mean , and two corrections due to waiting; Proposition 2.1 rewrites the corrections.
Formalization Note The identity is pathwise and uses no property of the waiting times; and are probability measures on .
Preamble
import Mathlib import Definitions.Def_ReedGGN_SystemEq_QueueLength
Formal statement
namespace ReedGGN.SystemEq
open MeasureTheory
/-- Equation (2.7) (p. 7): the number in system (2.2) decomposes as
`Q(t) = I(t) + W₀(t) + M₂(t) + A_G(t) + Σ_{i=1}^{(Q₀−N)⁺} (G(t − w̃_i) − G(t))
+ Σ_{i=1}^{A(t)} (G(t − τ_i − w_i) − G(t − τ_i))`. -/
theorem equation_2_7 (μ μ₀ : Measure ℝ) [IsProbabilityMeasure μ] [IsProbabilityMeasure μ₀]
(P : SamplePath) (t : ℝ) (ht : 0 ≤ t) :
(Q P t : ℝ) =
I μ μ₀ P t + W0 μ₀ P t + M2 μ P t + AG μ P t +
(∑ i ∈ Finset.Icc 1 (P.Q₀ - P.N), (G μ (t - P.wt i) - G μ t)) +
∑ i ∈ Finset.Icc 1 (P.A t), (G μ (t - P.τ i - P.w i) - G μ (t - P.τ i)) := by sorry
end ReedGGN.SystemEq
Source
Reed, The G/GI/N Queue in the Halfin–Whitt Regime, arXiv:0912.2837v1, p. 7, Eq. (2.7) (via (2.3))
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.