§3.1, p. 719 — conditional M/M/s waiting time is exponential
ProvedWhittEfficiency.MMs.sec_3_1_conditional_exponentialexponential-distributionmmcp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1queueing
Let be an integer and consider a stable, steady-state queue with arrival rate , unit service rate at each server, and stationary probabilities . Put , and let be the FCFS waiting time of an arriving customer. Then a positive wait has positive probability and
Thus the waiting time of a delayed customer has mean . This identifies the conditional distribution needed for Proposition 3.1.
Formalization Note The waiting-time law is the Erlang mixture defined in WhittEfficiency.MMs.WaitLaw. The conclusion is equality of the restriction of that measure to with its positive-wait mass times the exponential measure.
Preamble
import Mathlib import Definitions.Def_WhittEfficiency_MMs_WaitLaw open MeasureTheory ProbabilityTheory
Formal statement
namespace WhittEfficiency.MMs
/-- Whitt (1992), §3.1, p. 719: conditional on a positive wait, the M/M/s wait
is exponential with mean `1 / (s * (1 - ρ))`. -/
theorem sec_3_1_conditional_exponential
(s : ℕ) (hs : 1 ≤ s) (lam : ℝ) (hlam : 0 < lam) (hρ : lam < s)
(p : ℕ → ℝ)
(hp : QueueingFundamentals.BirthDeath.IsSteadyState
(fun _ => lam) (QueueingFundamentals.BirthDeath.mmcDeath 1 s) p) :
0 < (waitLaw s p).real (Set.Ioi (0 : ℝ)) ∧
(waitLaw s p).restrict (Set.Ioi (0 : ℝ)) =
(waitLaw s p) (Set.Ioi (0 : ℝ)) •
expMeasure ((s : ℝ) * (1 - lam / (s : ℝ))) := by sorry
end WhittEfficiency.MMs
Source
Whitt, Understanding the efficiency of multi-server service systems, Management Sci. 38 (1992), p. 719, §3.1, sentence immediately preceding Proposition 3.1; https://doi.org/10.1287/mnsc.38.5.708
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.