Beyond heavy-traffic regimes: Universal bounds and controls for the single-server (M/GI/1+GI) queue
|
|
- Cynthia Lester
- 5 years ago
- Views:
Transcription
1 1 / 33 Beyond heavy-traffic regimes: Universal bounds and controls for the single-server (M/GI/1+GI) queue Itai Gurvich Northwestern University Junfei Huang Chinese University of Hong Kong Stochastic Networks 216
2 2 / 33 The intuitive Derivation of a Brownian Queue Service,, 1 Load The waiting time/workload process in the M/GI/1 queue: A(t) W (t) = W () + s i (t I(t)) i=1 A(t) = W () (1 ρ)t (t I(t)) + s i ρt i=1 (M/GI/1) Ŵ (t) = W () (1 ρ)t (t Î(t)) + E[s 2 ]B(t) (Brownian Queue)
3 3 / 33 Brownian approximations as a model A tractable and useful tool in the modeler s toolbox Pricing in queues (e.g. Kim and Randhawa, 215) Competition between queues (e.g. Allon and Federgruen, 28) Contracting in services (e.g. Akan et. al. 211) Inventory Optimization (e.g. Allon and Van Mieghem 21) Initial Model difficult Brownian approximation? Accurate
4 Example: Dual Sourcing (Inventory) Net Inventory evolution (Mexico and.3 China renewal inputs S c, S M ): I(t) = I() + S c (t) +.25 S M (TM s (t)) D(t), 121 t TM s (t) = 1{I(u) <.2 s}ds. 9 zed via Simulation 5 Scaled cost: Brownian vs. optimal TBS Relative China cost c C /c M Scaled cost Relative China cost c C /c M.35 Optimization by simulation.3 = 1 = 1 = 1 Scaled cost Asymptotic (analytical).25 * C.2 Simulated cost of Brownian prescription = 1 = 1 = 1 Optimization by simulation = 1 = 1 = 1 Asymptotic (analytical) C * Relative China cost c C /c M, the relative error in scaled cost ption and the optimal control was ain implication is that the Browngood and useful approximation of and the cost Ĉ ˆ p ˆ M p when using the s formulae. One can observe that the scaled square Gad Allon, Jan Simulated A. Van cost Mieghem of Brownian (21). prescription Global Dual Sourcing: Tailored Base-Surge Allocation to Near- and Offshore = 1Production. = 1 Management = 1 Science 56(1): / 33
5 Example: Admission Control to a Many-Server Queue Koçağa, Yaşar Levent and Ward, Amy R (21). Admission control for a multi-server queue with abandonment. Queueing Systems, 65(3): ( ) 1 Asymptotic optimality for ρ() = 1 + O as (and N) Punchline: Universally accurate 5 / 33
6 Utilization (regime) assumptions and consequences 11 1 Service exp 1 Load 1.1 Patience exp 1 Embedding Consequence (as, µ ) ρ() = W ( ) Reflected OU (critical load) ρ() 1.1 (W ( ) µ (ρ() 1)) free OU θ (overload) Universal process approx in Ward and Glynn (23), Ward (212) 6 / 33
7 7 / 33 Sensitivity of the limit to patience modeling Patience hazard rate Mandelbaum Avishai and Sergey Zeltyn. (213) Data-stories about (im)patient customers in tele-queues. Queueing Systems 75(2), With ρ() = 1 β (critical load): E[W ] = O(1/ )
8 8 / 33 Sensitivity of the limit to patience modeling Consider the critically loaded M/M/1 + GI: ρ() = 1 β Finite patience drawn from a distribution F a ( ). Limit Theorems differ by model F a F does not scale with and has f a () > : diffusion limit has linear drift; Ward and Glynn (25); F a has hazard rate that scales with : F a (x) = 1 e x h( u)du, limit has non-linear drift; Reed and Ward (28).
9 9 / 33 Sensitivity of scaling to patience modeling Suppose ρ() = 1 (critical loading). F a =exponential (fixed): E[W ] = Θ ( ) 1, as. F a (x) = x 2 for x [, 1] (fixed): ( ) 1 E[W ] = Θ 1/3, as. Different patience dist. different scaling needed for limits. A result that bypasses case-by-case analysis and interpretation..
10 The M/GI/1 + GI queue Service, 1 Load Patience The virtual wait V (t) is the time an infinitely patient customer, arriving at time t, would have to wait. The waiting time is the minimum of the virtual wait and the customer s patience: W (t) = min(ν, V (t)). The first order ( fluid ) proxy for the stationary virtual wait is w that solves µ = F( w). 1 / 33
11 11 / 33 Dynamics of the virtual waiting time V (t) is the work contained in jobs that will not abandon. A(t) V (t) = V () + s i 1 {vi >ω i } (t I(t)). i=1 Satisfies the natural positivity properties, - V (t), t ; - I( ) is nondecreasing with I() = ; - 1 {V (s)>} di(s) =. Process limits for the GI/GI/1+GI queue: Ward and Glynn (23,25), Reed and Ward (28), Jennings and Reed (212)
12 The intuitive Brownian queue A(t) V (t) = V () + s i 1 {vi >ω i } (t I(t)) i=1 t = V () + + I(t) A(t) ρ F a (V (s))ds t + i=1 t s i 1 {vi >ω i } ρ F a (V (s))ds t V (t) = V () + ρ F a ( V (s))ds t + σb(t) + (µ Î(t), σ = )E[s1 2] ( x (dx) = G exp 2 π V ) ρ F a (u) 1 σ 2 du dx, x [, ). No scaling. The recommendation is to use π V as a proxy for π V. 12 / 33
13 A notion of approximation accuracy For the M/M/1 queue V = W and Ŵ (t) is a one dimensional RBM. If ρ < 1, Ŵ := Ŵ ( ) is expo(mean = ρ/(µ(1 ρ))) M/M/1 : E[W k ] = k!ρ (µ(1 ρ)) k, Brownian Q : E[Ŵ k ] = k!ρ k (µ(1 ρ)) k. The approximation gap for the k th moment is E[W k ] E[Ŵ k ] = ρk! (µ(1 ρ)) k (1 ρk 1 ) = k(1 ρk 1 ) ρ k 3 (1 ρ) E[ V k 1 ] k(k 1) ρ k 3 E[Ŵ k 1 ]. For k = 1 the gap is (the P-K formula for the M/GI/1 queue). 13 / 33
14 A notion of approximation accuracy For the M/GI/1 + GI queue, it universally holds E[W k ] E[Ŵ k ] C E[Ŵ k 1 ] Waiting Time Scaled Error Waiting Time Scaled Error Figure: Hyper Exponential patience: F a (x) = 4 7 (1 e 4x ) (1 e x/2 ). M/M/1+GI moments using Zeltyn and Mandelbaum (25). 14 / 33
15 15 / 33 Queue families The M/GI/1+GI queue primitives are p = (Arrival rate, Service time dist. F s, Patience dist. F a ) We will define Q-families parameterized by a constant H. and prove results of the form E[Wp k ] E[Ŵ p k ] sup C H p Q(H) E[Ŵ p k 1 ] Universality = size the family Q(H) Recall: scaling is sensitive to patience dist. and other primitives.
16 Queue family Q(H) = {p = (, F s, F a )} (i) service-time moments: E [ ( )] s 1 exp δ H E[s 1 ] H, and there exists a concentration constant c p (µh) 1 such that (ii) finite load: ρ [H 1, H], ρ 1 H. c p (iii) polynomial growth: F a is differentiable with density f a : (iv) concentration: and f a (y) H c 2 p ( 1 + y w p c p H), ρ F a (y) 1 H 1 1 c p, for all y w p + c p H, ρ F a (y) 1 H 1 1 c p, for all y w p c p H, Notice: c p, w p vary with the primitives. 16 / 33
17 Indeed a large family exp(θ) patience (ρ > 1 : ρe θ w = 1): ( ρ F a w + H ) 1 H 1 1 c p ( ) ρe θ w+ H 1 H 1 1 F a c p H ρ 1 ρ > 1 Infinite 1 (1 ρ) exp(θ) 1 (1 ρ) 1 1 Uniform[, α] 1 (1 ρ) 1 1 HyperExp(θ, θ) 1 (1 ρ) 1 1 Power(α, k) 1 (1 ρ) 1 k+1 k (ρ 1) 1 k 1 Erlang(k, θ) 1 (1 ρ) 1 k+1 k (ρ 1) 1 k 1 Beta(α, β) 1 (1 ρ) 1 α+1 α /ρ (ρ 1) 1 1 α max(θ, 2/θ, ρ, 1/ρ) max(1/α, α/ρ, ρ, 1/ρ) max(θ, 2/θ, ρ, 1/ρ) 1 α 1 ρ 2k (k ρ) (1 α) k 2 k+1 (ρ k)h E Γ(k) max(θ k, 1 ρ 1 θ k ) U 2 α+β (ρ α)γ(α+β) (ρ 1)Γ(α)Γ(β) min(l,1) U Table: c p and H for a family of patience distributions. All these are in one queue family. 17 / 33
18 18 / 33 The accuracy of the Brownian approximation Theorem (Virtual waiting time) Given H > and k N, there exists a constant C 1 H,k > such that E[(V w) k ] E[( V w) k ] = ± C1 H,k E[ V w k 1 ], p Q(H). Corollary (Waiting time) Given H and k N, there exists a constant C 2 H,k > such that E[W k ] E[Ŵ k ] = ± C2 H,k E[Ŵ k 1 ], p Q(H). For k = 1, the error is O(1/).
19 19 / 33 The accuracy of the Brownian approximation Corollary (Queue length) Given H, there exists a constant CH,1 2 > such that E[Q] = E[W ] = E[Ŵ ] ± C2 H,1, p Q(H). The mean-queue approximation gap is a constant. Corollary (Abandonment) Given H, there exists a constant C H > such that Ab = E[F a ( V )] ± C H 2 E[ V w 2 ], p Q(H). For example, if F a = exp(θ), the Ab approximation gap is O(1/).
20 2 / 33 About the tightness of the Q-family conditions Virtual Waiting Time Scaled Error M/D/1 + GI with ρ = 1 ( w = ) and F a = Gamma(.5, 2). F a = Gamma(.5, 2) violates our conditions.
21 21 / 33 c p captures concentration/scaling Lemma (Concentration bounds) There exist constants CH,k V, cv H,k > such that c V H,k E[ V w k ] c p k C V H,k, p Q(H). There exist C H,k, C 1 H,k > such that for all p Q(H), E[(V w) k ] E[( V w) k ] = ± C1 H,k E[ V w k 1 ] = ± C H,k ck 1 p.
22 c p captures concentration/scaling Lemma (Concentration bounds) There exist constants CH,k V, cv H,k > such that c V H,k E[ V w k ] c p k C V H,k, p Q(H). There exist C H,k, C 1 H,k > such that for all p Q(H), E[(V w) k ] E[( V w) k ] = ± C1 H,k E[ V w k 1 ] = ± C H,k ck 1 p. Example: F a = exp(θ) c p = 1 and E[(V w) k ] E[( V w) k ] = ± C H,k ( ) 1. k 1 21 / 33
23 22 / 33 Underlying Math: Generator comparisons (B&D) g k (x) = (x w)k E[( V w) k ] (E[ V w ]) k so that E[g k ( V )] = Solve (for Ψ) (ÂΨ)(x) = g k(x) (Brownian Poisson Eqn) E[(A Ψ)(V ) g k (V )] = E[(A Ψ)(V ) (Â Ψ)(V )] + E[(Â Ψ)(V ) g k (V )] = E[(A Ψ)(V ) (Â Ψ)(V )] If E[A Ψ(V )] = (Glynn and Zeevi (28)), then E[g k (V )] = E[(V w)k E[( V w) k ]] (E[ V w ]) k = E[(A Ψ)(V ) (Â Ψ)(V )]
24 Generator comparison and gradient bounds E[(V w) k ] E[( V w) k ] (E[ V w ]) k E[(A Ψ)(V ) (Â Ψ)(V )] A Ψ(x) = Ψ (1) (x) + F a (x) E [ Ψ(x + s1 ) Ψ(x) ] [ = Ψ (1) (x) + F a (x) E Ψ (1) (x)s ] 2 Ψ (2) (x)s1 2 + ɛ(x, s 1) = Â Ψ(x) + F a (x) E [ɛ(x, s 1 )]. E[ɛ(x, s 1 )] has Ψ s derivative of order m > 2. E[ A Ψ(V ) Â Ψ(V ) ] E[ F a (V ) E [ɛ(v, s 1 )]] show C H E[ V w ]. Where does V on the right-hand side come from? Show = Gradient + Apriori moment bounds (via c p drift cond.) See Braverman and Dai (216) 23 / 33
25 From performance analysis to optimization 24 / 33
26 Two ways in which regimes arise Consider a sequence of queues with ρ() = 1 β Identifying the optimal regime: Minimizing capacity + linear delay cost in the M/M/1 queue µ () := min µ c s µ + c w E[W (µ)] = + so that (1 ρ ()) = 1 + ρ cw 1 () 1 c s c w c s cw c s c w c s, as If c w = 1 4 and c s = 1, then, ρ = 1/2. 25 / 33
27 26 / 33 Dynamic optimization: Service-rate control in the M/G/1 queue Steps: Arrival rate ; Service time distribution F s with E[s 1 ] = 1. Controlled service rate µ(θ) = (1 + θ). Holding cost hx m. p = (, h). [ Jp,m V, 1 t ( = inf lim θ Θ V t t E x h(v (θ, s)) m + (θ(s)) 2) ] ds. An unscaled Brownian Control Problem (BCP) Universality over Q(H) = {(, h) : H 1, h (, H)} We will be agnostic to whether (or not) h scales down with
28 27 / 33 The Brownian control problem t t V (θ, t) =V () θ(s)ds + (1 + θ(s))1{v (θ, s) = }ds A(t) + s i t i=1 t V (θ, t) =V () θ(s)ds + + E[s1 2]B(t). t (1 + θ(s))1{ V (θ, s) = }ds [ J V, 1 t ( p,m = inf lim t t E x h( V (θ, s)) m + (θ(s)) 2) ] ds θ Θ V (BCP)
29 Universal optimality gap Theorem An optimal stationary (Brownian) policy θ p,m(x) exists and, for any p Q(H) := {(, h) : H 1, h (, H]}, J V, p,m J V p,m(θ p,m) B H (, m)j V, p,m 1 B H (, m) as. The gap is if m = / 33
30 Universal optimality gap Theorem An optimal stationary (Brownian) policy θ p,m(x) exists and, for any p Q(H) := {(, h) : H 1, h (, H]}, J V, p,m J V p,m(θ p,m) B H (, m)j V, p,m 1 B H (, m) as. The gap is if m = 2. Recall, we found for the (uncontrolled) M/GI/1 + GI queue: E[(V w) k ] E[( V w) k ] = ± C1 H,k E[ V w k 1 ] 28 / 33
31 For control, too, generator comparisons { γ = min z (Âz Ψ)(x) + (z) 2 + hx m}, (Diffusion HJB) Ψ() = Ψ (1) () = and Ψ (1) (x), for all x, Given ( Ψ, γ) : optimal service rate ẑ (x) = Ψ (1) (x) 2 γ = min z {(Az Ψ)(x) + (z)2 + hx m } (M/GI/1 Bellman) (Doshi (1978)) If for relevant values of z, A z Ψ Âz Ψ: min z {(Az Ψ)(x) + z + hx m } min z (Âz Ψ)(x) + z + hx m } Ψ, γ p,m almost solves the M/G/1 Bellman equation. 29 / 33
32 3 / 33 From BCP to M/GI/1 optimality Lemma Fix (p, m) and let ( Ψ, γ) be the solution the (BCPs) HJB equation. Then, for any admissible control θ for the M/G/1 queue (and any x, t ): E x [ t ( h(v (θ, s)) m + (θ(s)) 2) ] ds Ψ(x) ] E x [ Ψ(V (θ, t)) + γt [ t ( where A x (θ, t) = E x A θ(s) + A x (θ, t), ) ] Ψ(V (θ, s)) Âθ(s) Ψ(V (θ, s)) ds. If θ is the BCP stationary control ẑ, the inequality is replaced with equality. If m = 2, A x (θ, t) for any control θ.
33 31 / 33 From BCP to M/GI/1 optimality Lemma Fix m and let ( Ψ p,m, γ p,m ) be the (family of) solutions to the HJB equation. Then, there exist constants CH,m 1, C2 H,m such that, for any order optimal family of policies {θ p,m, p Q(H)}, lim inf t 1 t Ax p,m(θ p,m, t) C 1 H,m BH (, m)j Y, p,m 1, x, and under the stationary policy θ p,m, lim sup t 1 t Ax p,m(θ p,m, t) C 2 H,m BH (, m)j Y, p,m 1, x.
34 32 / 33 Conclusion In great generality, it is fine to use the intuitive Brownian queue of the M/GI/1+GI. It is universally accurate in regime and patience-scaling. Similar ideas are applied to static and dynamic optimization Underlying math: Avoid scaling through Q-families. From the universal proximity of operators to the universal proximity of equation solutions (Poisson or HJB).
35 33 / 33 Time Dependent Expectations First moment Second moment Scaled gap Time Time Figure: Time-dependent performance for M/M/1 with µ = 1: (LHS) ρ =.9, (RHS) Scaled gap
Designing a Telephone Call Center with Impatient Customers
Designing a Telephone Call Center with Impatient Customers with Ofer Garnett Marty Reiman Sergey Zeltyn Appendix: Strong Approximations of M/M/ + Source: ErlangA QEDandSTROG FIAL.tex M/M/ + M System Poisson
More informationOverflow Networks: Approximations and Implications to Call-Center Outsourcing
Overflow Networks: Approximations and Implications to Call-Center Outsourcing Itai Gurvich (Northwestern University) Joint work with Ohad Perry (CWI) Call Centers with Overflow λ 1 λ 2 Source of complexity:
More informationDynamic Control of Parallel-Server Systems
Dynamic Control of Parallel-Server Systems Jim Dai Georgia Institute of Technology Tolga Tezcan University of Illinois at Urbana-Champaign May 13, 2009 Jim Dai (Georgia Tech) Many-Server Asymptotic Optimality
More informationClass 11 Non-Parametric Models of a Service System; GI/GI/1, GI/GI/n: Exact & Approximate Analysis.
Service Engineering Class 11 Non-Parametric Models of a Service System; GI/GI/1, GI/GI/n: Exact & Approximate Analysis. G/G/1 Queue: Virtual Waiting Time (Unfinished Work). GI/GI/1: Lindley s Equations
More informationThe M/M/n+G Queue: Summary of Performance Measures
The M/M/n+G Queue: Summary of Performance Measures Avishai Mandelbaum and Sergey Zeltyn Faculty of Industrial Engineering & Management Technion Haifa 32 ISRAEL emails: avim@tx.technion.ac.il zeltyn@ie.technion.ac.il
More informationStabilizing Customer Abandonment in Many-Server Queues with Time-Varying Arrivals
OPERATIONS RESEARCH Vol. 6, No. 6, November December 212, pp. 1551 1564 ISSN 3-364X (print) ISSN 1526-5463 (online) http://dx.doi.org/1.1287/opre.112.114 212 INFORMS Stabilizing Customer Abandonment in
More informationDesigning load balancing and admission control policies: lessons from NDS regime
Designing load balancing and admission control policies: lessons from NDS regime VARUN GUPTA University of Chicago Based on works with : Neil Walton, Jiheng Zhang ρ K θ is a useful regime to study the
More informationA Diffusion Approximation for Stationary Distribution of Many-Server Queueing System In Halfin-Whitt Regime
A Diffusion Approximation for Stationary Distribution of Many-Server Queueing System In Halfin-Whitt Regime Mohammadreza Aghajani joint work with Kavita Ramanan Brown University APS Conference, Istanbul,
More informationStein s Method for Steady-State Approximations: Error Bounds and Engineering Solutions
Stein s Method for Steady-State Approximations: Error Bounds and Engineering Solutions Jim Dai Joint work with Anton Braverman and Jiekun Feng Cornell University Workshop on Congestion Games Institute
More informationQueues with Many Servers and Impatient Customers
MATHEMATICS OF OPERATIOS RESEARCH Vol. 37, o. 1, February 212, pp. 41 65 ISS 364-765X (print) ISS 1526-5471 (online) http://dx.doi.org/1.1287/moor.111.53 212 IFORMS Queues with Many Servers and Impatient
More informationLecture 9: Deterministic Fluid Models and Many-Server Heavy-Traffic Limits. IEOR 4615: Service Engineering Professor Whitt February 19, 2015
Lecture 9: Deterministic Fluid Models and Many-Server Heavy-Traffic Limits IEOR 4615: Service Engineering Professor Whitt February 19, 2015 Outline Deterministic Fluid Models Directly From Data: Cumulative
More informationThe Generalized Drift Skorokhod Problem in One Dimension: Its solution and application to the GI/GI/1 + GI and. M/M/N/N transient distributions
The Generalized Drift Skorokhod Problem in One Dimension: Its solution and application to the GI/GI/1 + GI and M/M/N/N transient distributions Josh Reed Amy Ward Dongyuan Zhan Stern School of Business
More informationErlang-C = M/M/N. agents. queue ACD. arrivals. busy ACD. queue. abandonment BACK FRONT. lost calls. arrivals. lost calls
Erlang-C = M/M/N agents arrivals ACD queue Erlang-A lost calls FRONT BACK arrivals busy ACD queue abandonment lost calls Erlang-C = M/M/N agents arrivals ACD queue Rough Performance Analysis
More informationClassical Queueing Models.
Sergey Zeltyn January 2005 STAT 99. Service Engineering. The Wharton School. University of Pennsylvania. Based on: Classical Queueing Models. Mandelbaum A. Service Engineering course, Technion. http://iew3.technion.ac.il/serveng2005w
More informationElectronic Companion Fluid Models for Overloaded Multi-Class Many-Server Queueing Systems with FCFS Routing
Submitted to Management Science manuscript MS-251-27 Electronic Companion Fluid Models for Overloaded Multi-Class Many-Server Queueing Systems with FCFS Routing Rishi Talreja, Ward Whitt Department of
More informationRouting and Staffing in Large-Scale Service Systems: The Case of Homogeneous Impatient Customers and Heterogeneous Servers 1
Routing and Staffing in Large-Scale Service Systems: The Case of Homogeneous Impatient Customers and Heterogeneous Servers 1 Mor Armony 2 Avishai Mandelbaum 3 June 25, 2008 Abstract Motivated by call centers,
More informationMaximum pressure policies for stochastic processing networks
Maximum pressure policies for stochastic processing networks Jim Dai Joint work with Wuqin Lin at Northwestern Univ. The 2011 Lunteren Conference Jim Dai (Georgia Tech) MPPs January 18, 2011 1 / 55 Outline
More informationMASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.070J Fall 2013 Lecture 22 12/09/2013. Skorokhod Mapping Theorem. Reflected Brownian Motion
MASSACHUSETTS INSTITUTE OF TECHNOLOGY 6.265/15.7J Fall 213 Lecture 22 12/9/213 Skorokhod Mapping Theorem. Reflected Brownian Motion Content. 1. G/G/1 queueing system 2. One dimensional reflection mapping
More informationLIMITS AND APPROXIMATIONS FOR THE M/G/1 LIFO WAITING-TIME DISTRIBUTION
LIMITS AND APPROXIMATIONS FOR THE M/G/1 LIFO WAITING-TIME DISTRIBUTION by Joseph Abate 1 and Ward Whitt 2 April 15, 1996 Revision: January 2, 1997 Operations Research Letters 20 (1997) 199 206 1 900 Hammond
More informationStochastic Networks and Parameter Uncertainty
Stochastic Networks and Parameter Uncertainty Assaf Zeevi Graduate School of Business Columbia University Stochastic Processing Networks Conference, August 2009 based on joint work with Mike Harrison Achal
More informationQueueing Theory I Summary! Little s Law! Queueing System Notation! Stationary Analysis of Elementary Queueing Systems " M/M/1 " M/M/m " M/M/1/K "
Queueing Theory I Summary Little s Law Queueing System Notation Stationary Analysis of Elementary Queueing Systems " M/M/1 " M/M/m " M/M/1/K " Little s Law a(t): the process that counts the number of arrivals
More informationProactive Care with Degrading Class Types
Proactive Care with Degrading Class Types Yue Hu (DRO, Columbia Business School) Joint work with Prof. Carri Chan (DRO, Columbia Business School) and Prof. Jing Dong (DRO, Columbia Business School) Motivation
More informationRouting and Staffing in Large-Scale Service Systems: The Case of Homogeneous Impatient Customers and Heterogeneous Servers
OPERATIONS RESEARCH Vol. 59, No. 1, January February 2011, pp. 50 65 issn 0030-364X eissn 1526-5463 11 5901 0050 informs doi 10.1287/opre.1100.0878 2011 INFORMS Routing and Staffing in Large-Scale Service
More informationThe Performance Impact of Delay Announcements
The Performance Impact of Delay Announcements Taking Account of Customer Response IEOR 4615, Service Engineering, Professor Whitt Supplement to Lecture 21, April 21, 2015 Review: The Purpose of Delay Announcements
More informationDesign, staffing and control of large service systems: The case of a single customer class and multiple server types. April 2, 2004 DRAFT.
Design, staffing and control of large service systems: The case of a single customer class and multiple server types. Mor Armony 1 Avishai Mandelbaum 2 April 2, 2004 DRAFT Abstract Motivated by modern
More informationStabilizing Customer Abandonment in Many-Server Queues with Time-Varying Arrivals
Submitted to Operations Research manuscript OPRE-29-6-259; OPRE-211-12-597.R1 Stabilizing Customer Abandonment in Many-Server Queues with -Varying Arrivals Yunan Liu Department of Industrial Engineering,
More informationAsymptotic Coupling of an SPDE, with Applications to Many-Server Queues
Asymptotic Coupling of an SPDE, with Applications to Many-Server Queues Mohammadreza Aghajani joint work with Kavita Ramanan Brown University March 2014 Mohammadreza Aghajanijoint work Asymptotic with
More informationSTAFFING A CALL CENTER WITH UNCERTAIN ARRIVAL RATE AND ABSENTEEISM
STAFFING A CALL CENTER WITH UNCERTAIN ARRIVAL RATE AND ABSENTEEISM by Ward Whitt Department of Industrial Engineering and Operations Research Columbia University, New York, NY 10027 6699 Abstract This
More informationOther properties of M M 1
Other properties of M M 1 Přemysl Bejda premyslbejda@gmail.com 2012 Contents 1 Reflected Lévy Process 2 Time dependent properties of M M 1 3 Waiting times and queue disciplines in M M 1 Contents 1 Reflected
More informationTechnical Appendix for: When Promotions Meet Operations: Cross-Selling and Its Effect on Call-Center Performance
Technical Appendix for: When Promotions Meet Operations: Cross-Selling and Its Effect on Call-Center Performance In this technical appendix we provide proofs for the various results stated in the manuscript
More informationTechnical Appendix for: When Promotions Meet Operations: Cross-Selling and Its Effect on Call-Center Performance
Technical Appendix for: When Promotions Meet Operations: Cross-Selling and Its Effect on Call-Center Performance In this technical appendix we provide proofs for the various results stated in the manuscript
More informationTopics in queueing theory
Graduate Theses and Dissertations Iowa State University Capstones, Theses and Dissertations 217 Topics in queueing theory Keguo Huang Iowa State University Follow this and additional works at: http://lib.dr.iastate.edu/etd
More informationarxiv: v1 [math.pr] 21 Jan 2014
Diffusion Models for Double-ended Queues with Renewal Arrival Processes arxiv:141.5146v1 [math.pr] 21 Jan 214 Xin Liu 1, Qi Gong 2, and Vidyadhar G. Kulkarni 2 1 Department of Mathematical Sciences, Clemson
More informationBlind Fair Routing in Large-Scale Service Systems with Heterogeneous Customers and Servers
OPERATIONS RESEARCH Vol. 6, No., January February 23, pp. 228 243 ISSN 3-364X (print) ISSN 526-5463 (online) http://dx.doi.org/.287/opre.2.29 23 INFORMS Blind Fair Routing in Large-Scale Service Systems
More informationAuthor's personal copy
Queueing Syst (215) 81:341 378 DOI 1.17/s11134-15-9462-x Stabilizing performance in a single-server queue with time-varying arrival rate Ward Whitt 1 Received: 5 July 214 / Revised: 7 May 215 / Published
More informationEconomy of Scale in Multiserver Service Systems: A Retrospective. Ward Whitt. IEOR Department. Columbia University
Economy of Scale in Multiserver Service Systems: A Retrospective Ward Whitt IEOR Department Columbia University Ancient Relics A. K. Erlang (1924) On the rational determination of the number of circuits.
More informationManaging Call Centers with Many Strategic Agents
Managing Call Centers with Many Strategic Agents Rouba Ibrahim, Kenan Arifoglu Management Science and Innovation, UCL YEQT Workshop - Eindhoven November 2014 Work Schedule Flexibility 86SwofwthewqBestwCompanieswtowWorkwForqwofferw
More informationIEOR 8100: Topics in OR: Asymptotic Methods in Queueing Theory. Fall 2009, Professor Whitt. Class Lecture Notes: Wednesday, September 9.
IEOR 8100: Topics in OR: Asymptotic Methods in Queueing Theory Fall 2009, Professor Whitt Class Lecture Notes: Wednesday, September 9. Heavy-Traffic Limits for the GI/G/1 Queue 1. The GI/G/1 Queue We will
More informationAPPENDIX to Stabilizing Performance in Many-Server Queues with Time-Varying Arrivals and Customer Feedback
APPENDIX to Stabilizing Performance in Many-Server Queues with -Varying Arrivals and Customer Feedback Yunan Liu and Ward Whitt Department of Industrial Engineering North Carolina State University Raleigh,
More informationState Space Collapse in Many-Server Diffusion Limits of Parallel Server Systems. J. G. Dai. Tolga Tezcan
State Space Collapse in Many-Server Diffusion imits of Parallel Server Systems J. G. Dai H. Milton Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, Georgia
More informationTwo Workload Properties for Brownian Networks
Two Workload Properties for Brownian Networks M. Bramson School of Mathematics University of Minnesota Minneapolis MN 55455 bramson@math.umn.edu R. J. Williams Department of Mathematics University of California,
More informationControl of Many-Server Queueing Systems in Heavy Traffic. Gennady Shaikhet
Control of Many-Server Queueing Systems in Heavy Traffic Gennady Shaikhet Control of Many-Server Queueing Systems in Heavy Traffic Research Thesis In Partial Fulfillment of the Requirements for the Degree
More informationSENSITIVITY OF PERFORMANCE IN THE ERLANG-A QUEUEING MODEL TO CHANGES IN THE MODEL PARAMETERS
SENSITIVITY OF PERFORMANCE IN THE ERLANG-A QUEUEING MODEL TO CHANGES IN THE MODEL PARAMETERS by Ward Whitt Department of Industrial Engineering and Operations Research Columbia University, New York, NY
More informationBIRTH DEATH PROCESSES AND QUEUEING SYSTEMS
BIRTH DEATH PROCESSES AND QUEUEING SYSTEMS Andrea Bobbio Anno Accademico 999-2000 Queueing Systems 2 Notation for Queueing Systems /λ mean time between arrivals S = /µ ρ = λ/µ N mean service time traffic
More informationAdvanced Computer Networks Lecture 3. Models of Queuing
Advanced Computer Networks Lecture 3. Models of Queuing Husheng Li Min Kao Department of Electrical Engineering and Computer Science University of Tennessee, Knoxville Spring, 2016 1/13 Terminology of
More informationModeling Service Networks with Time-Varying Demand
Modeling Service with Time-Varying Demand - Performance Approximations and Staffing Controls (with Beixiang He, Liam Huang, Korhan Aras, Ward Whitt) Department of Industrial and Systems Engineering NC
More informationStochastic grey-box modeling of queueing systems: fitting birth-and-death processes to data
Queueing Syst (2015) 79:391 426 DOI 10.1007/s11134-014-9429-3 Stochastic grey-box modeling of queueing systems: fitting birth-and-death processes to data James Dong Ward Whitt Received: 8 January 2014
More informationM/G/1 and Priority Queueing
M/G/1 and Priority Queueing Richard T. B. Ma School of Computing National University of Singapore CS 5229: Advanced Compute Networks Outline PASTA M/G/1 Workload and FIFO Delay Pollaczek Khinchine Formula
More informationMaking Delay Announcements
Making Delay Announcements Performance Impact and Predicting Queueing Delays Ward Whitt With Mor Armony, Rouba Ibrahim and Nahum Shimkin March 7, 2012 Last Class 1 The Overloaded G/GI/s + GI Fluid Queue
More informationPoint Process Control
Point Process Control The following note is based on Chapters I, II and VII in Brémaud s book Point Processes and Queues (1981). 1 Basic Definitions Consider some probability space (Ω, F, P). A real-valued
More informationBlind Fair Routing in Large-Scale Service Systems with Heterogeneous Customers and Servers
Blind Fair Routing in Large-Scale Service Systems with Heterogeneous Customers and Servers Mor Armony Amy R. Ward 2 October 6, 2 Abstract In a call center, arriving customers must be routed to available
More informationAsymptotics for Polling Models with Limited Service Policies
Asymptotics for Polling Models with Limited Service Policies Woojin Chang School of Industrial and Systems Engineering Georgia Institute of Technology Atlanta, GA 30332-0205 USA Douglas G. Down Department
More informationA Robust Queueing Network Analyzer Based on Indices of Dispersion
A Robust Queueing Network Analyzer Based on Indices of Dispersion Wei You (joint work with Ward Whitt) Columbia University INFORMS 2018, Phoenix November 6, 2018 1/20 Motivation Many complex service systems
More informationManaging Service Systems with an Offline Waiting Option and Customer Abandonment: Companion Note
Managing Service Systems with an Offline Waiting Option and Customer Abandonment: Companion Note Vasiliki Kostami Amy R. Ward September 25, 28 The Amusement Park Ride Setting An amusement park ride departs
More informationCritically Loaded Time-Varying Multi-Server Queues: Computational Challenges and Approximations
Critically Loaded Time-Varying Multi-Server Queues: Computational Challenges and Approximations Young Myoung Ko Sabre Holdings, 315 Sabre Drive, Southlake, Texas 7692, USA, YoungMyoung.Ko@sabre.com Natarajan
More informationPositive Harris Recurrence and Diffusion Scale Analysis of a Push Pull Queueing Network. Haifa Statistics Seminar May 5, 2008
Positive Harris Recurrence and Diffusion Scale Analysis of a Push Pull Queueing Network Yoni Nazarathy Gideon Weiss Haifa Statistics Seminar May 5, 2008 1 Outline 1 Preview of Results 2 Introduction Queueing
More informationHEAVY-TRAFFIC EXTREME-VALUE LIMITS FOR QUEUES
HEAVY-TRAFFIC EXTREME-VALUE LIMITS FOR QUEUES by Peter W. Glynn Department of Operations Research Stanford University Stanford, CA 94305-4022 and Ward Whitt AT&T Bell Laboratories Murray Hill, NJ 07974-0636
More informationDynamic Control of a Tandem Queueing System with Abandonments
Dynamic Control of a Tandem Queueing System with Abandonments Gabriel Zayas-Cabán 1 Jungui Xie 2 Linda V. Green 3 Mark E. Lewis 1 1 Cornell University Ithaca, NY 2 University of Science and Technology
More informationOn the Resource/Performance Tradeoff in Large Scale Queueing Systems
On the Resource/Performance Tradeoff in Large Scale Queueing Systems David Gamarnik MIT Joint work with Patrick Eschenfeldt, John Tsitsiklis and Martin Zubeldia (MIT) High level comments High level comments
More informationService Level Agreements in Call Centers: Perils and Prescriptions
Service Level Agreements in Call Centers: Perils and Prescriptions Joseph M. Milner Joseph L. Rotman School of Management University of Toronto Tava Lennon Olsen John M. Olin School of Business Washington
More informationStein s Method for Steady-State Approximations: A Toolbox of Techniques
Stein s Method for Steady-State Approximations: A Toolbox of Techniques Anton Braverman Based on joint work with Jim Dai (Cornell), Itai Gurvich (Cornell), and Junfei Huang (CUHK). November 17, 2017 Outline
More informationService Engineering January Laws of Congestion
Service Engineering January 2004 http://ie.technion.ac.il/serveng2004 Laws of Congestion The Law for (The) Causes of Operational Queues Scarce Resources Synchronization Gaps (in DS PERT Networks) Linear
More informationGideon Weiss University of Haifa. Joint work with students: Anat Kopzon Yoni Nazarathy. Stanford University, MSE, February, 2009
Optimal Finite Horizon Control of Manufacturing Systems: Fluid Solution by SCLP (separated continuous LP) and Fluid Tracking using IVQs (infinite virtual queues) Stanford University, MSE, February, 29
More informationQueueing Theory II. Summary. ! M/M/1 Output process. ! Networks of Queue! Method of Stages. ! General Distributions
Queueing Theory II Summary! M/M/1 Output process! Networks of Queue! Method of Stages " Erlang Distribution " Hyperexponential Distribution! General Distributions " Embedded Markov Chains M/M/1 Output
More informationThe Offered-Load Process: Modeling, Inference and Applications. Research Thesis
The Offered-Load Process: Modeling, Inference and Applications Research Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of Master of Science in Statistics Michael Reich Submitted
More informationec1 e-companion to Liu and Whitt: Stabilizing Performance
ec1 This page is intentionally blank. Proper e-companion title page, with INFORMS branding and exact metadata of the main paper, will be produced by the INFORMS office when the issue is being assembled.
More informationLIMITS FOR QUEUES AS THE WAITING ROOM GROWS. Bell Communications Research AT&T Bell Laboratories Red Bank, NJ Murray Hill, NJ 07974
LIMITS FOR QUEUES AS THE WAITING ROOM GROWS by Daniel P. Heyman Ward Whitt Bell Communications Research AT&T Bell Laboratories Red Bank, NJ 07701 Murray Hill, NJ 07974 May 11, 1988 ABSTRACT We study the
More informationAsymptotic Delay Distribution and Burst Size Impact on a Network Node Driven by Self-similar Traffic
Èíôîðìàöèîííûå ïðîöåññû, Òîì 5, 1, 2005, ñòð. 4046. c 2004 D'Apice, Manzo. INFORMATION THEORY AND INFORMATION PROCESSING Asymptotic Delay Distribution and Burst Size Impact on a Network Node Driven by
More informationStaffing Call-Centers With Uncertain Demand Forecasts: A Chance-Constrained Optimization Approach
Staffing Call-Centers With Uncertain Demand Forecasts: A Chance-Constrained Optimization Approach Itai Gurvich James Luedtke Tolga Tezcan May 25, 2009 Abstract We consider the problem of staffing large-scale
More informationLoad Balancing in Distributed Service System: A Survey
Load Balancing in Distributed Service System: A Survey Xingyu Zhou The Ohio State University zhou.2055@osu.edu November 21, 2016 Xingyu Zhou (OSU) Load Balancing November 21, 2016 1 / 29 Introduction and
More informationMarkovian N-Server Queues (Birth & Death Models)
Markovian -Server Queues (Birth & Death Moels) - Busy Perio Arrivals Poisson (λ) ; Services exp(µ) (E(S) = /µ) Servers statistically ientical, serving FCFS. Offere loa R = λ E(S) = λ/µ Erlangs Q(t) = number
More informationDesign heuristic for parallel many server systems under FCFS-ALIS Adan, I.J.B.F.; Boon, M.A.A.; Weiss, G.
Design heuristic for parallel many server systems under FCFS-ALIS Adan, I.J.B.F.; Boon, M.A.A.; Weiss, G. Published in: arxiv Published: 04/03/2016 Document Version Accepted manuscript including changes
More informationVARUN GUPTA. Takayuki Osogami (IBM Research-Tokyo) Carnegie Mellon Google Research University of Chicago Booth School of Business.
VARUN GUPTA Carnegie Mellon Google Research University of Chicago Booth School of Business With: Taayui Osogami (IBM Research-Toyo) 1 2 Homogeneous servers 3 First-Come-First-Serve Buffer Homogeneous servers
More informationIEOR 6711, HMWK 5, Professor Sigman
IEOR 6711, HMWK 5, Professor Sigman 1. Semi-Markov processes: Consider an irreducible positive recurrent discrete-time Markov chain {X n } with transition matrix P (P i,j ), i, j S, and finite state space.
More informationOnline Supplement: Managing Hospital Inpatient Bed Capacity through Partitioning Care into Focused Wings
Online Supplement: Managing Hospital Inpatient Bed Capacity through Partitioning Care into Focused Wings Thomas J. Best, Burhaneddin Sandıkçı, Donald D. Eisenstein University of Chicago Booth School of
More informationA How Hard are Steady-State Queueing Simulations?
A How Hard are Steady-State Queueing Simulations? ERIC CAO NI and SHANE G. HENDERSON, Cornell University Some queueing systems require tremendously long simulation runlengths to obtain accurate estimators
More informationDelay Announcements. Predicting Queueing Delays for Delay Announcements. IEOR 4615, Service Engineering, Professor Whitt. Lecture 21, April 21, 2015
Delay Announcements Predicting Queueing Delays for Delay Announcements IEOR 4615, Service Engineering, Professor Whitt Lecture 21, April 21, 2015 OUTLINE Delay Announcements: Why, What and When? [Slide
More informationQ = (c) Assuming that Ricoh has been working continuously for 7 days, what is the probability that it will remain working at least 8 more days?
IEOR 4106: Introduction to Operations Research: Stochastic Models Spring 2005, Professor Whitt, Second Midterm Exam Chapters 5-6 in Ross, Thursday, March 31, 11:00am-1:00pm Open Book: but only the Ross
More informationFluid Limit of A Many-Server Queueing Network with Abandonment
Fluid Limit of A Many-Server Queueing Networ with Abandonment Weining Kang Guodong Pang October 22, 213 Abstract This paper studies a non-marovian many-server queueing networ with abandonment, where externally
More informationSingle-Server Service-Station (G/G/1)
Service Engineering July 997 Last Revised January, 006 Single-Server Service-Station (G/G/) arrivals queue 000000000000 000000000000 departures Arrivals A = {A(t), t 0}, counting process, e.g., completely
More informationFair Dynamic Routing in Large-Scale Heterogeneous-Server Systems
OPERATIONS RESEARCH Vol. 58, No. 3, May June 2010, pp. 624 637 issn 0030-364X eissn 1526-5463 10 5803 0624 informs doi 10.1287/opre.1090.0777 2010 INFORMS Fair Dynamic Routing in Large-Scale Heterogeneous-Server
More informationAsymptotically Optimal Inventory Control For Assemble-to-Order Systems
Asymptotically Optimal Inventory Control For Assemble-to-Order Systems Marty Reiman Columbia Univerisity joint work with Mustafa Dogru, Haohua Wan, and Qiong Wang May 16, 2018 Outline The Assemble-to-Order
More informationDesign and evaluation of overloaded service systems with skill based routing, under FCFS policies
Design and evaluation of overloaded service systems with skill based routing, under FCFS policies Ivo Adan Marko Boon Gideon Weiss April 2, 2013 Abstract We study an overloaded service system with servers
More informationCDA5530: Performance Models of Computers and Networks. Chapter 4: Elementary Queuing Theory
CDA5530: Performance Models of Computers and Networks Chapter 4: Elementary Queuing Theory Definition Queuing system: a buffer (waiting room), service facility (one or more servers) a scheduling policy
More informationWhat You Should Know About Queueing Models To Set Staffing Requirements in Service Systems
What You Should Know About Queueing Models To Set Staffing Requirements in Service Systems by Ward Whitt Department of Industrial Engineering and Operations Research Columbia University 304 S. W. Mudd
More informatione - c o m p a n i o n
OPERATIONS RESEARCH http://dx.doi.org/1.1287/opre.111.13ec e - c o m p a n i o n ONLY AVAILABLE IN ELECTRONIC FORM 212 INFORMS Electronic Companion A Diffusion Regime with Nondegenerate Slowdown by Rami
More informationESTIMATING WAITING TIMES WITH THE TIME-VARYING LITTLE S LAW
ESTIMATING WAITING TIMES WITH THE TIME-VARYING LITTLE S LAW Song-Hee Kim and Ward Whitt Industrial Engineering and Operations Research Columbia University New York, NY, 10027 {sk3116, ww2040}@columbia.edu
More informationENGINEERING SOLUTION OF A BASIC CALL-CENTER MODEL
ENGINEERING SOLUTION OF A BASIC CALL-CENTER MODEL by Ward Whitt Department of Industrial Engineering and Operations Research Columbia University, New York, NY 10027 Abstract An algorithm is developed to
More informationOperations Research Letters. Instability of FIFO in a simple queueing system with arbitrarily low loads
Operations Research Letters 37 (2009) 312 316 Contents lists available at ScienceDirect Operations Research Letters journal homepage: www.elsevier.com/locate/orl Instability of FIFO in a simple queueing
More informationSince D has an exponential distribution, E[D] = 0.09 years. Since {A(t) : t 0} is a Poisson process with rate λ = 10, 000, A(0.
IEOR 46: Introduction to Operations Research: Stochastic Models Chapters 5-6 in Ross, Thursday, April, 4:5-5:35pm SOLUTIONS to Second Midterm Exam, Spring 9, Open Book: but only the Ross textbook, the
More informationA Note on the Event Horizon for a Processor Sharing Queue
A Note on the Event Horizon for a Processor Sharing Queue Robert C. Hampshire Heinz School of Public Policy and Management Carnegie Mellon University hamp@andrew.cmu.edu William A. Massey Department of
More informationGI/M/1 and GI/M/m queuing systems
GI/M/1 and GI/M/m queuing systems Dmitri A. Moltchanov moltchan@cs.tut.fi http://www.cs.tut.fi/kurssit/tlt-2716/ OUTLINE: GI/M/1 queuing system; Methods of analysis; Imbedded Markov chain approach; Waiting
More informationStaffing Many-Server Queues with Impatient Customers: Constraint Satisfaction in Call Centers
OPERATIONS RESEARCH Vol. 57, No. 5, September October 29, pp. 1189 125 issn 3-364X eissn 1526-5463 9 575 1189 informs doi 1.1287/opre.18.651 29 INFORMS Staffing Many-Server Queues with Impatient Customers:
More informationA Fluid Approximation for Service Systems Responding to Unexpected Overloads
OPERATIONS RESEARCH Vol. 59, No. 5, September October 2011, pp. 1159 1170 issn 0030-364X eissn 1526-5463 11 5905 1159 http://dx.doi.org/10.1287/opre.1110.0985 2011 INFORMS A Fluid Approximation for Service
More informationAnalysis of a Two-Phase Queueing System with Impatient Customers and Multiple Vacations
The Tenth International Symposium on Operations Research and Its Applications (ISORA 211) Dunhuang, China, August 28 31, 211 Copyright 211 ORSC & APORC, pp. 292 298 Analysis of a Two-Phase Queueing System
More informationStaffing many-server queues with impatient customers: constraint satisfaction in call centers
Staffing many-server queues with impatient customers: constraint satisfaction in call centers Avishai Mandelbaum Faculty of Industrial Engineering & Management, Technion, Haifa 32000, Israel, avim@tx.technion.ac.il
More informationThe shortest queue problem
The shortest queue problem Ivo Adan March 19, 2002 1/40 queue 1 join the shortest queue queue 2 Where: Poisson arrivals with rate Exponential service times with mean 1/ 2/40 queue 1 queue 2 randomly assign
More informationDynamic Matching Models
Dynamic Matching Models Ana Bušić Inria Paris - Rocquencourt CS Department of École normale supérieure joint work with Varun Gupta, Jean Mairesse and Sean Meyn 3rd Workshop on Cognition and Control January
More informationAbandonment and Customers Patience in Tele-Queues. The Palm/Erlang-A Model.
Sergey Zeltyn February 25 zeltyn@ie.technion.ac.il STAT 991. Service Engineering. The Wharton School. University of Pennsylvania. Abandonment and Customers Patience in Tele-Queues. The Palm/Erlang-A Model.
More informationNon Markovian Queues (contd.)
MODULE 7: RENEWAL PROCESSES 29 Lecture 5 Non Markovian Queues (contd) For the case where the service time is constant, V ar(b) = 0, then the P-K formula for M/D/ queue reduces to L s = ρ + ρ 2 2( ρ) where
More information