Article Contents
Article Contents

# Steady-state and first passage time distributions for waiting times in the MAP/M/s+G queueing model with generally distributed patience times

• * Corresponding author: Nail Akar
• We study the $MAP/M/s+G$ queueing model that arises in various multi-server engineering problems including telephone call centers, under the assumption of MAP (Markovian Arrival Process) arrivals, exponentially distributed service times, infinite waiting room, and generally distributed patience times. Using sample-path arguments, we propose to obtain the steady-state distribution of the virtual waiting time and subsequently the other relevant performance metrics of interest via the steady-state solution of a certain Continuous Feedback Fluid Queue (CFFQ). The proposed method is exact when the patience time is a discrete random variable and is asymptotically exact when it is continuous/hybrid, for which case discretization of the patience time distribution is required giving rise to a computational complexity depending linearly on the number of discretization levels. Additionally, a novel method is proposed to accurately obtain the first passage time distributions for the virtual and actual waiting times again using CFFQs while approximating the deterministic time horizons by Erlang distributions or more efficient Concentrated Matrix Exponential (CME) distributions. Numerical results are presented to validate the effectiveness of the proposed numerical method.

Mathematics Subject Classification: Primary: 60K30, 65C40, 68M20, 90B22; Secondary: 60J25.

 Citation:

• Figure 1.  Venn diagram illustrating the relationship between CMFQs, MRMFQs, and CFFQs

Figure 2.  Sample path of the random process $Z_f(t), t\geq 0$

Figure 3.  Sample path of the random process $\tilde{Z}_f(t), t\geq 0$

Figure 4.  Sample path of the random process $Z_f(t), t\geq 0$

Table 1.  Results for the $M/M/s+M$ scenario with $s=10$, $\lambda=10$, $\mu=1$, $E(A)=1$

 Analysis $K=10$ Analysis $K=50$ Analysis $K=250$ Ref. [59] Simulation Results $\Pr \left\{ {W=0 } \right\}$ 0.46000 0.45802 0.45794 0.45793 $0.45829 \pm 0.00209$ $\Pr \left\{ {W=0|\mathcal{S} } \right\}$ 0.52612 0.52353 0.52343 0.52341 $0.52375 \pm 0.00195$ $\Pr \left\{ {\mathcal{A}} \right\}$ 0.12567 0.12513 0.12511 0.12511 $0.12498 \pm 0.00084$ $E(W|\mathcal{S})$ 0.11648 0.11500 0.11494 0.11494 $0.11491 \pm 0.00077$ $\text{Var}(W|\mathcal{S})$ 0.03377 0.03310 0.03307 0.03307 $0.03299 \pm 0.00030$ $F_{W|\mathcal{S},W >0}(.1)$ 0.28108 0.28100 0.28108 0.28107 $0.28060 \pm 0.00093$ $F_{W|\mathcal{S},W >0}(.2)$ 0.51155 0.51285 0.51283 0.51283 $0.51208 \pm 0.00145$

Table 2.  Comparison of analytical results with simulations for the $MAP/M/s+M$ scenario

 Analysis $K=10$} Analysis $K=50$} Analysis $K=250$} Simulation Results $\Pr \left\{ { W=0 } \right\}$ 0.43620 0.43464 0.43458 $0.43473 \pm 0.00097$ $\Pr \left\{ { W=0|\mathcal{S}} \right\}$ 0.51361 0.51152 0.51143 $0.51158 \pm 0.00096$ $\Pr \left\{ { \mathcal{A} } \right\}$ 0.15073 0.15029 0.15027 $0.15023 \pm 0.00043$ $E(W|\mathcal{S})$ 0.14046 0.13743 0.13737 $0.13718 \pm 0.00043$ $\text{Var}(W|\mathcal{S})$ 0.12746 0.04456 0.04451 $0.04437 \pm 0.00018$ $F_{W|\mathcal{S},W >0}(.1)$ 0.23845 0.23847 0.23854 $0.23854 \pm 0.00066$ $F_{W|\mathcal{S},W >0}(.2)$ 0.44512 0.44644 0.44642 $0.44661 \pm 0.00094$

Table 3.  Comparison of analytical results with simulations for the $MAP/M/s+M_B$ scenario

 Analysis $K=10$} Analysis $K=50$} Analysis $K=250$} Simulation Results $\Pr \left\{ { W=0 } \right\}$ 0.69025 0.68969 0.68968 $0.68977 \pm 0.00034$ $\Pr \left\{ { W=0|\mathcal{S}} \right\}$ 0.88516 0.88428 0.88425 $0.88430 \pm 0.00019$ $\Pr \left\{ { \mathcal{A} } \right\}$ 0.22020 0.22006 0.22006 $0.21999 \pm 0.00024$ $E(W|\mathcal{S})$ 0.01563 0.01483 0.01480 $0.01481 \pm 0.00004$ $\text{Var}(W|\mathcal{S})$ 0.00368 0.00344 0.00343 $0.00344 \pm 0.00001$ $F_{W|\mathcal{S},W>0}(.1)$ 0.53344 0.53371 0.53384 $0.53332\pm 0.00061$ $F_{W|\mathcal{S},W>0}(.2)$ 0.78709 0.78846 0.78846 $0.78832\pm 0.00055$

Table 4.  Comparison of analytical results with simulations for the $MAP/M/s+HE$ scenario

 Analysis $K=10$} Analysis $K=50$} Analysis $K=250$} Simulation Results $\Pr \left\{ { W=0 } \right\}$ 0.39314 0.39104 0.39096 $0.39136 \pm 0.00088$ $\Pr \left\{ { W=0|\mathcal{S}} \right\}$ 0.45551 0.45276 0.45266 $0.45311 \pm 0.00087$ $\Pr \left\{ { \mathcal{A} } \right\}$ 0.13692 0.13632 0.13630 $0.13629 \pm 0.00037$ $E(W|\mathcal{S})$ 0.20596 0.20195 0.20180 $0.20175 \pm 0.00061$ $\text{Var}(W|\mathcal{S})$ 0.08667 0.08281 0.08266 $0.08266 \pm 0.00039$ $F_{W|\mathcal{S},W>0}(.1)$ 0.18329 0.18628 0.18639 $0.18610\pm 0.00054$ $F_{W|\mathcal{S},W>0}(.2)$ 0.35371 0.35537 0.35547 $0.35501\pm 0.00085$

Table 5.  Comparison of analytical results with simulations for the $MAP/M/s+D$ scenario

 Analysis Simulation Results $\Pr \left\{ { W=0} \right\}$ 0.37989 $0.38040 \pm 0.00091$ $\Pr \left\{ { W=0|\mathcal{S}} \right\}$ 0.43851 $0.43893 \pm 0.00090$ $\Pr \left\{ { \mathcal{A} } \right\}$ 0.13367 $0.13334 \pm 0.00042$ $E(W|\mathcal{S})$ 0.14990 $0.14974 \pm 0.00033$ $\text{Var}(W|\mathcal{S})$ 0.02964 $0.02962 \pm 0.00004$ $F_{W|\mathcal{S},W>0}(.1)$ 0.17763 $0.17778 \pm 0.00047$ $F_{W|\mathcal{S},W>0}(.2)$ 0.35825 $0.35855 \pm 0.00072$

Table 6.  Comparison of analytical results with simulations for the $MAP/M/s+W$ scenario

 Analysis $K=10$ Analysis $K=50$ Analysis $K=250$ Simulation Results $\Pr \left\{ { W=0 } \right\}$ 0.34776 0.33324 0.33138 $0.33166 \pm 0.00097$ $\Pr \left\{ { W=0|\mathcal{S}} \right\}$ 0.39620 0.37774 0.37538 $0.37557 \pm 0.00098$ $\Pr \left\{ { \mathcal{A} } \right\}$ 0.12227 0.11778 0.11721 $0.11693 \pm 0.00042$ $E(W|\mathcal{S})$ 0.31926 0.25150 0.25138 $0.25104 \pm 0.00065$ $\text{Var}(W|\mathcal{S})$ 3.91308 0.12642 0.08071 $0.07974 \pm 0.00018$ $F_{W|\mathcal{S},W>0}(.1)$ 0.13862 0.13371 0.13361 $0.13401 \pm 0.00046$ $F_{W|\mathcal{S},W>0}(.2)$ 0.27149 0.26618 0.26670 $0.26723 \pm 0.00070$

Table 7.  Comparison of analytical results with simulations for the $MAP/M/s+E_2$ scenario

 Analysis $K=10$ Analysis $K=50$ Analysis $K=250$ Simulation Results $\Pr \left\{ { W=0 } \right\}$ 0.30664 0.29490 0.29380 $0.29336 \pm 0.00115$ $\Pr \left\{ { W=0|\mathcal{S}} \right\}$ 0.34379 0.32928 0.32793 $0.32737 \pm 0.00115$ $\Pr \left\{ { \mathcal{A} } \right\}$ 0.10806 0.10442 0.10408 $0.10393 \pm 0.00048$ $E(W|\mathcal{S})$ 0.37254 0.37450 0.37482 $0.37475 \pm 0.00138$ $\text{Var}(W|\mathcal{S})$ 0.17112 0.17499 0.17560 $0.17527 \pm 0.00075$ $F_{W|\mathcal{S},W>0}(.1)$ 0.109761 0.10744 0.10785 $0.10782 \pm 0.00047$ $F_{W|\mathcal{S},W>0}(.2)$ 0.21430 0.21334 0.21375 $0.21371 \pm 0.00082$

Table 8.  Comparison of analytical results with simulations for the $MAP/M/s+E_3$ scenario

 Analysis $K=10$ Analysis $K=50$ Analysis $K=250$ Simulation Results $\Pr \left\{ { W=0 } \right\}$ 0.24365 0.22187 0.21925 $0.21946 \pm 0.00106$ $\Pr \left\{ { W=0|\mathcal{S}} \right\}$ 0.26659 0.24087 0.23780 $0.23791 \pm 0.00108$ $\Pr \left\{ { \mathcal{A} } \right\}$ 0.08603 0.07886 0.07801 $0.07757 \pm 0.00044$ $E(W|\mathcal{S})$ 0.63951 0.65123 0.65284 $0.65137 \pm 0.00222$ $\text{Var}(W|\mathcal{S})$ 0.34263 0.38303 0.38981 $0.39046 \pm 0.00157$ $F_{W|\mathcal{S},W>0}(.1)$ 0.07563 0.06845 0.06783 $0.06795 \pm 0.00035$ $F_{W|\mathcal{S},W>0}(.2)$ 0.14729 0.13513 0.13433 $0.13458 \pm 0.00061$

Table 9.  The cdf of the virtual waiting time evaluated at time $\tau$, $F_v^{a,b,\pi_0,\theta_0}(\tau)$, obtained by the proposed method using Erlang-$\ell$ and CME-$\ell$ distributions for three values of the parameter $\ell$, and for the scenario $\theta_0 = [\begin{smallmatrix} 1 & 0 \end{smallmatrix}]$

 Erlang-$\ell$ CME-$\ell$ $\tau$ $b$ Simulation $\ell=25$ $\ell=51$ $\ell=101$ $\ell=25$ $\ell=51$ $\ell=101$ $1/4$ 0.61906$\pm$0.00027 0.59251 0.60529 0.61200 0.61705 0.61883 0.61922 $1/2$ 0.39325$\pm$0.00050 0.38744 0.39032 0.39185 0.39309 0.39345 0.39354 $1$ $1$ 0.11791$\pm$0.00065 0.13245 0.12551 0.12189 0.11922 0.11822 0.11800 $2$ 0.00323$\pm$0.00044 0.00607 0.00456 0.00388 0.00345 0.00328 0.00324 $4$ 0.00000$\pm$0.00002 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 $1/4$ 0.97700$\pm$0.00022 0.97692 0.97698 0.97700 0.97690 0.97699 0.97700 $1/2$ 0.96964$\pm$0.00029 0.96910 0.96949 0.96960 0.96953 0.96965 0.96967 $5$ $1$ 0.95339$\pm$0.00031 0.94898 0.95163 0.95257 0.95294 0.95326 0.95331 $2$ 0.86128$\pm$0.00039 0.82707 0.84443 0.85283 0.85859 0.86077 0.86124 $4$ 0.14516$\pm$0.00050 0.16177 0.15408 0.14986 0.14667 0.14544 0.14517 $1/4$ 0.98695$\pm$0.00016 0.98684 0.98691 0.98694 0.98693 0.84307 0.98697 $1/2$ 0.98274$\pm$0.00020 0.98261 0.98270 0.98274 0.98273 0.98278 0.98278 $25$ $1$ 0.97408$\pm$0.00022 0.97381 0.97394 0.97400 0.97399 0.97405 0.97406 $2$ 0.95078$\pm$0.00035 0.95051 0.95073 0.95084 0.95084 0.95093 0.95094 $4$ 0.86036$\pm$0.00035 0.85903 0.85979 0.86006 0.86011 0.86027 0.86030

Table 10.  The cdf of the virtual waiting time evaluated at time $\tau$, $F_v^{a,b,\pi_0,\theta_0}(\tau)$, obtained by the proposed method using Erlang-$\ell$ and CME-$\ell$ distributions for three values of the parameter $\ell$, and for the scenario $\theta_0 = [\begin{smallmatrix} 0 & 1 \end{smallmatrix}]$

 Erlang-$\ell$ CME-$\ell$ $\tau$ $b$ Simulation $\ell=25$ $\ell=51$ $\ell=101$ $\ell=25$ $\ell=51$ $\ell=101$ $1/4$ 0.00663$\pm$0.00010 0.00703 0.00681 0.00671 0.00664 0.00661 0.00660 $1/2$ 0.00355$\pm$0.00006 0.00400 0.00377 0.00365 0.00357 0.00354 0.00353 $1$ $1$ 0.00081$\pm$0.00003 0.00111 0.00096 0.00089 0.00084 0.00082 0.00082 $2$ 0.00002$\pm$0.00001 0.00004 0.00002 0.00002 0.00002 0.00002 0.00002 $4$ 0.00000$\pm$0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 $1/4$ 0.11115$\pm$0.00032 0.11067 0.11085 0.11094 0.11100 0.11102 0.11103 $1/2$ 0.10346$\pm$0.00031 0.10305 0.10323 0.10332 0.10338 0.10340 0.10341 $5$ $1$ 0.08827$\pm$0.00026 0.08794 0.08809 0.08816 0.08823 0.08824 0.08825 $2$ 0.05324$\pm$0.00020 0.05382 0.05349 0.05334 0.05326 0.05322 0.05321 $4$ 0.00365$\pm$0.00005 0.00538 0.00453 0.00412 0.00382 0.00372 0.00369 $1/4$ 0.49669$\pm$0.00059 0.49161 0.49415 0.49537 0.49620 0.49653 0.49660 $1/2$ 0.48996$\pm$0.00058 0.48497 0.48751 0.48872 0.48956 0.48989 0.48996 $25$ $1$ 0.47661$\pm$0.00058 0.47159 0.47410 0.47531 0.47614 0.47647 0.47654 $2$ 0.44349$\pm$0.00059 0.43858 0.44104 0.44223 0.44304 0.44336 0.44343 $4$ 0.33702$\pm$0.00049 0.33222 0.33441 0.33547 0.33622 0.33649 0.33654

Table 11.  The cdf of the actual waiting time evaluated at time $\tau$, $F_a^{a,b,\pi_0,\theta_0}(\tau)$, obtained by the proposed method using Erlang-$\ell$ and CME-$\ell$ distributions for three values of the parameter $\ell$, and for the scenario $\theta_0 = [\begin{smallmatrix} 1 & 0 \end{smallmatrix}]$

 Erlang-$\ell$ CME-$\ell$ $\tau$ $b$ Simulation $\ell=25$ $\ell=51$ $\ell=101$ $\ell=25$ $\ell=51$ $\ell=101$ $1/4$ 0.52092$\pm$0.00051 0.50227 0.51121 0.51598 0.51964 0.52089 0.52117 $1/2$ 0.31345$\pm$0.00049 0.31514 0.31438 0.31397 0.31375 0.31358 0.31355 $1$ $1$ 0.08271$\pm$0.00028 0.09792 0.09062 0.08685 0.08408 0.08305 0.08282 $2$ 0.00184$\pm$0.00004 0.00382 0.00273 0.00226 0.00197 0.00186 0.00183 $4$ 0.00000$\pm$0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 $1/4$ 0.97418$\pm$0.00020 0.97393 0.97406 0.97409 0.97400 0.97409 0.97411 $1/2$ 0.96674$\pm$0.00023 0.96574 0.96635 0.96652 0.96649 0.96662 0.96664 $5$ $1$ 0.94941$\pm$0.00023 0.94345 0.94692 0.94819 0.94876 0.94915 0.94922 $2$ 0.84264$\pm$0.00052 0.80496 0.82380 0.83310 0.83960 0.84204 0.84256 $4$ 0.11830$\pm$0.00040 0.13624 0.12782 0.12323 0.11975 0.11844 0.11815 $1/4$ 0.98533$\pm$0.00015 0.98517 0.98524 0.98528 0.98527 0.98530 0.98531 $1/2$ 0.98108$\pm$0.00018 0.98089 0.98098 0.98103 0.98102 0.98106 0.98107 $25$ $1$ 0.97205$\pm$0.00017 0.97180 0.97193 0.97200 0.97199 0.97205 0.97206 $2$ 0.94819$\pm$0.00020 0.94767 0.94789 0.94800 0.94801 0.94810 0.94811 $4$ 0.85268$\pm$0.00038 0.85115 0.85203 0.85234 0.85240 0.85257 0.85260

Table 12.  The cdf of the actual waiting time evaluated at time $\tau$, $F_a^{a,b,\pi_0,\theta_0}(\tau)$, obtained by the proposed method using Erlang-$\ell$ and CME-$\ell$ distributions for three values of the parameter $\ell$, and for the scenario $\theta_0 = [\begin{smallmatrix} 0 & 1 \end{smallmatrix}]$

 Erlang-$\ell$ CME-$\ell$ $\tau$ $b$ Simulation $\ell=25$ $\ell=51$ $\ell=101$ $\ell=25$ $\ell=51$ $\ell=101$ $1/4$ 0.00502$\pm$0.00005 0.00553 0.00529 0.00517 0.00509 0.00506 0.00505 $1/2$ 0.00257$\pm$0.00004 0.00307 0.00283 0.00271 0.00263 0.00260 0.00260 $1$ $1$ 0.00053$\pm$0.00003 0.00078 0.00066 0.00060 0.00056 0.00054 0.00054 $2$ 0.00001$\pm$0.00000 0.00002 0.00002 0.00001 0.00001 0.00001 0.00001 $4$ 0.00000$\pm$0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 0.00000 $1/4$ 0.10802$\pm$0.00030 0.10761 0.10780 0.10789 0.10795 0.10797 0.10798 $1/2$ 0.10030$\pm$0.00029 0.10000 0.10018 0.10026 0.10033 0.10035 0.10035 $5$ $1$ 0.08481$\pm$0.00026 0.08463 0.08476 0.08483 0.08489 0.08491 0.08491 $2$ 0.04944$\pm$0.00022 0.05027 0.04985 0.04966 0.04954 0.04949 0.04948 $4$ 0.00283$\pm$0.00007 0.00438 0.00362 0.00324 0.00298 0.00289 0.00287 $1/4$ 0.49362$\pm$0.00051 0.48896 0.49149 0.49271 0.49355 0.49387 0.49394 $1/2$ 0.48698$\pm$0.00051 0.48230 0.48483 0.48605 0.48688 0.48721 0.48727 $25$ $1$ 0.47332$\pm$0.00051 0.46860 0.47111 0.47232 0.47315 0.47347 0.47354 $2$ 0.43941$\pm$0.00051 0.43475 0.43720 0.43839 0.43920 0.43951 0.43958 $4$ 0.32884$\pm$0.00048 0.32438 0.32654 0.32759 0.32832 0.32859 0.32865
•  [1] S. Ahn, A. L. Badescu and V. Ramaswami, Time dependent analysis of finite buffer fluid flows and risk models with a dividend barrier, Queueing Syst., 55 (2007), 207-222.  doi: 10.1007/s11134-007-9017-x. [2] S. Ahn and V. Ramaswami, Efficient algorithms for transient analysis of stochastic fluid flow models, Journal of Applied Probability, 42 (2005), 531-549.  doi: 10.1239/jap/1118777186. [3] N. Akar, O. Gursoy, G. Horvath and M. Telek, Transient and First Passage Time Distributions of First- and Second-Order Multi-Regime Markov Fluid Queues via ME-fication, accepted for publication in Methodology and Computing in Applied Probability, 2020. [4] N. Akar and K. Sohraby, Infinite- and finite-buffer Markov fluid queues: A unified analysis, J. Appl. Probab., 41 (2004), 557-569.  doi: 10.1239/jap/1082999086. [5] Z. Aksin, M. Armony and V. Mehrotra, The modern call center: A multi-disciplinary perspective on operations management research, Production and Operations Management, 16 (2007), 665-688.  doi: 10.1111/j.1937-5956.2007.tb00288.x. [6] A. T. Andersen and B. F. Nielsen, A Markovian approach for modeling packet traffic with long-range dependence, IEEE Journal on Selected Areas in Communications, 16 (1998), 719-732.  doi: 10.1109/49.700908. [7] D. Anick, D. Mitra and M. M. Sondhi, Stochastic theory of a data-handling system with multiple sources, Bell System Technical Journal, 61 (1982), 1871-1894.  doi: 10.1002/j.1538-7305.1982.tb03089.x. [8] S. Asmussen and M. Bladt, Renewal theory and queueing algorithms for matrix-exponential distributions, in Matrix-Analytic Methods in Stochastic Models, Lecture Notes in Pure and Appl. Math., 183, Dekker, New York, 1997,313–341. [9] S. Asmussen and G. Koole, Marked point processes as limits of Markovian arrival streams, Journal of Applied Probability, 30 (1993), 365-372.  doi: 10.2307/3214845. [10] S. Asmussen and J. R. Møller, Calculation of the steady state waiting time distribution in GI/PH/c and MAP/PH/c queues, Queueing Systems, 37 (2001), 9-29.  doi: 10.1023/A:1011083915877. [11] F. Baccelli and G. Hebuterne, On queues with impatient customers, in Performance '81, (Amsterdam, 1981), North-Holland, Amsterdam-New York, 1981., 159–179. [12] A. Badescu, S. Drekic and D. Landriault, On the analysis of a multi-threshold Markovian risk model, Scandinavian Actuarial Journal, 2007 (2007), 248-260.  doi: 10.1080/03461230701554080. [13] N. G. Bean and B. F. Nielsen, Quasi-birth-and-death processes with rational arrival process components, Stochastic Models, 26 (2010), 309-334.  doi: 10.1080/15326349.2010.498311. [14] N. G. Bean and M. M. O'Reilly, Performance measures of a multi-layer Markovian fluid model, Annals of Operations Research, 160 (2008), 99-120.  doi: 10.1007/s10479-007-0299-5. [15] A. Brandt and M. Brandt, On the M(n)/M(n)/s queue with impatient calls, Performance Evaluation, 35 (1999), 1-18. [16] A. Brandt and M. Brandt, Asymptotic results and a Markovian approximation for the M(n)/M(n)/s+GI system, Queueing Systems, 41 (2002), 73-94.  doi: 10.1023/A:1015781818360. [17] L. Brown, N. Gans, A. Mandelbaum, A. Sakov, H. Shen, S. Zeltyn and L. Zhao, Statistical analysis of a telephone call center, Journal of the American Statistical Association, 100 (2005), 36-50.  doi: 10.1198/016214504000001808. [18] P. Buchholz and M. Telek, Stochastic Petri nets with matrix exponentially distributed firing times, Performance Evaluation, 67 (2010), 1373-1385.  doi: 10.1016/j.peva.2010.08.023. [19] P. Buchholz and M. Telek, Rational processes related to communicating Markov processes, Journal of Applied Probability, 49 (2012), 40-59.  doi: 10.1239/jap/1331216833. [20] G. Casale, Building accurate workload models using Markovian arrival processes, in Proceedings of the ACM SIGMETRICS Joint International Conference on Measurement and Modeling of Computer Systems, SIGMETRICS '11, ACM, New York, NY, USA, (2011), 357–358. doi: 10.1145/1993744.1993783. [21] G. Casale, E. Z. Zhang and E. Smirni, KPC-Toolbox: Best recipes for automatic trace fitting using Markovian Arrival Processes, Performance Evaluation, 67 (2010), 873-896.  doi: 10.1016/j.peva.2009.12.003. [22] B. D. Choi, B. Kim and D. Zhu, MAP/M/c Queue with Constant Impatient Time, Mathematics of Operations Research, 29 (2004), 309-325.  doi: 10.1287/moor.1030.0081. [23] A. da Silva Soares and G. Latouche, Fluid queues with level dependent evolution, European Journal of Operational Research, 196 (2009), 1041-1048.  doi: 10.1016/j.ejor.2008.05.010. [24] M. W. Fackrell, Characterization of Matrix-Exponential Distributions, PhD thesis, The University of Adelaide, 2003. [25] N. Gans, G. Koole and A. Mandelbaum, Telephone call centers: Tutorial, review, and research prospects, Manufacturing & Service Operations Management, 5 (2003), 79-141.  doi: 10.1287/msom.5.2.79.16071. [26] O. Garnett, A. Mandelbaum and M. Reiman, Designing a call center with impatient customers, Manufacturing & Service Operations Management, 4 (2002), 208-227.  doi: 10.1287/msom.4.3.208.7753. [27] Q.-M. He and H. Wu, Multi-layer MMFF processes and the MAP/PH/K+GI queue: Theory and algorithms, Queueing Models and Service Management, 3 (2020), 37-87. [28] Q.-M. He and H. Zhang, On matrix exponential distributions, Advances in Applied Probability, 39 (2007), 271-292.  doi: 10.1239/aap/1175266478. [29] G. Horváth, I. Horváth and M. Telek, High order concentrated matrix-exponential distributions, Stochastic Models, 36 (2020), 176-192.  doi: 10.1080/15326349.2019.1702058. [30] I. Horváth, O. Sáfár, M. Telek and B. Zámbó, Concentrated matrix exponential distributions, in Computer Performance Engineering (eds. D. Fiems, M. Paolieri and A. N. Platis), Springer International Publishing, Cham, (2016), 18–31. [31] O. Jouini, G. Koole and A. Roubos, Performance indicators for call centers with impatient customers, IIE Transactions, 45 (2013), 341-354.  doi: 10.1080/0740817X.2012.712241. [32] H. E. Kankaya and N. Akar, Solving multi-regime feedback fluid queues, Stochastic Models, 24 (2008), 425-450.  doi: 10.1080/15326340802232285. [33] K. Kawanishi and T. Takine, MAP/M/c and M/PH/c queues with constant impatience times, Queueing Systems, 82 (2016), 381-420.  doi: 10.1007/s11134-015-9455-9. [34] C. Kim, S. Dudin, O. Taramin and J. Baek, Queueing system $MAP|PH|N|N+R$ with impatient heterogeneous customers as a model of call center, Applied Mathematical Modelling, 37 (2013), 958-976.  doi: 10.1016/j.apm.2012.03.021. [35] C. Knessl and J. S. H. van Leeuwaarden, Transient analysis of the Erlang A model, Mathematical Methods of Operations Research, 82 (2015), 143-173.  doi: 10.1007/s00186-015-0498-9. [36] R. Kumar and S. Sharma, Transient analysis of an M/M/c queuing system with balking and retention of reneging customers, Communications in Statistics - Theory and Methods, 47 (2018), 1318-1327.  doi: 10.1080/03610926.2017.1319485. [37] G. Latouche and V. Ramaswami, Introduction to Matrix Analytical Methods in Stochastic Modeling, ASA-SIAM Series on Statistics and Applied Probability. Society for Industrial and Applied Mathematics (SIAM), Philadelphia, PA; American Statistical Association, Alexandria, VA, 1999 doi: 10.1137/1.9780898719734. [38] D. M. Lucantoni, New results for the single server queue with a batch Markovian arrival process, Comm. Statist. Stochastic Models, 7 (1991), 1-46.  doi: 10.1080/15326349108807174. [39] D. M. Lucantoni, K. S. Meier-Hellstern and M. F. Neuts, A single server queue with server vacations and a class of nonrenewal arrival processes, Adv. Applied Prob., 22 (1990), 676-705.  doi: 10.2307/1427464. [40] A. Mandelbaum and S. Zeltyn, The impact of customers' patience on delay and abandonment: Some empirically-driven experiments with the $M/M/ n + G$ queue, OR Spectrum, 26 (2004), 377-411.  doi: 10.1007/s00291-004-0164-8. [41] M. Mandjes, D. Mitra and W. Scheinhardt, Models of network access using feedback fluid queues, Queueing Syst., 44 (2003), 365-398.  doi: 10.1023/A:1025147422141. [42] K. Mitchell, Constructing a correlated sequence of matrix exponentials with invariant first-order properties, Operations Research Letters, 28 (2001), 27-34.  doi: 10.1016/S0167-6377(00)00062-6. [43] A. Movaghar, On queueing with customer impatience until the beginning of service, Queueing Systems, 29 (1998), 337-350.  doi: 10.1023/A:1019196416987. [44] M. F. Neuts, Matrix-geometric Solutions in Stochastic Models: An Algorithmic Approach, Dover Publications, Inc., 1981. [45] M. F. Neuts, Structured Stochastic Matrices of M/G/1 Type and Their Applications, Marcel Dekker, N.Y., 1989. [46] H. Okamura, T. Dohi and K. S. Trivedi, Markovian arrival process parameter estimation with group data, IEEE/ACM Trans. Netw., 17 (2009), 1326-1339.  doi: 10.1109/TNET.2008.2008750. [47] C. Palm, Methods of judging the annoyance caused by congestion, Tele, 4 (1953), 189-208. [48] W. Scheinhardt, N. Van Foreest and M. Mandjes, Continuous feedback fluid queues, Operations Research Letters, 33 (2005), 551-559.  doi: 10.1016/j.orl.2004.11.008. [49] H. C. Tijms, Stochastic Modelling and Analysis: A Computational Approach, John Wiley & Sons, Ltd., Chichester, 1986. [50] R. C. F. Tucker, Accurate method for analysis of a packet-speech multiplexer with limited delay, IEEE Transactions on Communications, 36 (1988), 479-483.  doi: 10.1109/26.2773. [51] B. Van Houdt, Analysis of the adaptive MMAP[K]/PH[K]/1 queue: A multi-type queue with adaptive arrivals and general impatience, European Journal of Operational Research, 220 (2012), 695-704.  doi: 10.1016/j.ejor.2012.02.029. [52] B. Van Houdt and C. Blondia, Approximated transient queue length and waiting time distributions via steady state analysis, Stochastic Models, 21 (2005), 725-744.  doi: 10.1081/STM-200056027. [53] J. Van Velthoven, B. Van Houdt and C. Blondia, Simultaneous transient analysis of QBD Markov Chains for all initial configurations using a level based recursion, in Fourth International Conference on the Quantitative Evaluation of Systems (QEST 2007), (2007), 79–90. doi: 10.1109/QEST.2007.21. [54] W. Whitt, Improving service by informing customers about anticipated delays, Management Science, 45 (1999), 192-207.  doi: 10.1287/mnsc.45.2.192. [55] W. Whitt, Engineering solution of a basic call-center model, Management Science, 51 (2005), 221-235.  doi: 10.1287/mnsc.1040.0302. [56] M. A. Yazici and N. Akar, The workload-dependent MAP/PH/1 queue with infinite/finite workload capacity, Performance Evaluation, 70 (2013), 1047-1058.  doi: 10.1016/j.peva.2013.09.002. [57] M. A. Yazici and N. Akar, The finite/infinite horizon ruin problem with multi-threshold premiums: a Markov fluid queue approach, Annals of Operations Research, 252 (2017), 85-99.  doi: 10.1007/s10479-015-2105-0. [58] S. Zelytn, Call Centers with Impatient Customers: Exact Analysis and Many-Server Asymptotics of the $M/M/n+G$ Queue, Phd, Technion - Israel Institute of Technology, 2004. [59] S. Zeltyn and A. Mandelbaum, Call centers with impatient customers: Many-server asymptotics of the $M/M/n + G$ queue, Queueing Systems, 51 (2005), 361-402.  doi: 10.1007/s11134-005-3699-8.

Figures(4)

Tables(12)