Skip to main content
Log in

On two-queue Markovian polling systems with exhaustive service

  • Published:
Queueing Systems Aims and scope Submit manuscript

Abstract

We consider a class of two-queue polling systems with exhaustive service, where the order in which the server visits the queues is governed by a discrete-time Markov chain. For this model, we derive an expression for the probability generating function of the joint queue length distribution at polling epochs. Based on these results, we obtain explicit expressions for the Laplace–Stieltjes transforms of the waiting-time distributions and the probability generating function of the joint queue length distribution at an arbitrary point in time. We also study the heavy-traffic behaviour of properly scaled versions of these distributions, which results in compact and closed-form expressions for the distribution functions themselves. The heavy-traffic behaviour turns out to be similar to that of cyclic polling models, provides insights into the main effects of the model parameters when the system is heavily loaded, and can be used to derive closed-form approximations for the waiting-time distribution or the queue length distribution.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

References

  1. Abate, J., Whitt, W.: Numerical inversion of probability generating functions. Oper. Res. Lett. 12, 245–251 (1992)

    Article  Google Scholar 

  2. Abate, J., Whitt, W.: Numerical inversion of Laplace transforms of probability distributions. ORSA J. Comput. 7, 36–43 (1995)

    Article  Google Scholar 

  3. Athreya, K.B., Ney, P.E.: Branching Processes. Springer, New York (1972)

    Book  Google Scholar 

  4. Boon, M.A.A., van der Mei, R.D., Winands, E.M.M.: Applications of polling systems. Surv. Oper. Res. Manag. Sci. 16, 67–82 (2011)

    Google Scholar 

  5. Boon, M.A.A., Winands, E.M.M.: Heavy-traffic analysis of \(k\)-limited polling systems. Probab. Eng. Inf. Sci. (2014, to appear)

  6. Boon, M.A.A., Winands, E.M.M., Adan, I.J.B.F., van Wijk, A.C.C.: Closed-form waiting time approximations for polling systems. Perform. Eval. 68, 290–306 (2011)

    Article  Google Scholar 

  7. Boxma, O.J., Groenendijk, W.P.: Two queues with alternating service and switching times. In: Boxma, O.J., Syski, R. (eds.) Queueing Theory and its Applications (Liber Amicorum for J. W. Cohen), pp. 261–282. North-Holland, Amsterdam (1988)

    Google Scholar 

  8. Boxma, O.J., Kella, O., Kosiński, K.M.: Queue lengths and workloads in polling systems. Oper. Res. Lett. 39, 401–405 (2011)

    Article  Google Scholar 

  9. Boxma, O.J., Weststrate, J.: Waiting times in polling systems with Markovian server routing. In: Stiege, G., Lie, J.S. (eds.) Messung, Modellierung und Bewertung von Rechensystemen und Netzen, pp. 89–104. Springer, Berlin (1989)

    Chapter  Google Scholar 

  10. Coffman, E.G., Puhalskii, A.A., Reiman, M.I.: Polling systems with zero switch-over times: a heavy-traffic principle. Ann. Appl. Probab. 5, 681–719 (1995)

    Article  Google Scholar 

  11. Coffman, E.G., Puhalskii, A.A., Reiman, M.I.: Polling systems in heavy-traffic: a Bessel process limit. Math. Oper. Res. 23, 257–304 (1998)

    Article  Google Scholar 

  12. den Iseger, P.: Numerical transform inversion using Gaussian quadrature. Probab. Eng. Inf. Sci. 20, 1–44 (2006)

    Google Scholar 

  13. Dorsman, J.L., Borst, S.C., Boxma, O.J., Vlasiou, M.: Markovian polling systems with an application to wireless random-access networks. Technical Report 2014–001, Eurandom Preprint Series, 2014. http://www.eurandom.tue.nl/reports/ (2014)

  14. Dorsman, J.L., van der Mei, R.D., Winands, E.M.M.: A new method for deriving waiting-time approximations in polling systems with renewal arrivals. Stoch. Models 27, 318–332 (2011)

    Article  Google Scholar 

  15. Grossglauser, M., Tse, D.: Mobility increases the capacity of ad hoc wireless networks. IEEE/ACM Trans. Netw. 10, 477–486 (2002)

    Article  Google Scholar 

  16. Holma, H., Toskala, A.: HSDPA/HSUPA for UMTS: High Speed Radio Access for Mobile Communications. Wiley, Hoboken, NJ (2006)

    Book  Google Scholar 

  17. Keilson, J., Servi, L.D.: The distributional form of Little’s law and the Fuhrmann–Cooper decomposition. Oper. Res. Lett. 9, 239–247 (1990)

    Article  Google Scholar 

  18. Kleinrock, L., Levy, H.: The analysis of random polling systems. Oper. Res. 36, 716–732 (1988)

    Article  Google Scholar 

  19. Konheim, A.G., Levy, H., Srinivasan, M.M.: Descendant set: an efficient approach for the analysis of polling systems. IEEE Trans. Commun. 42(234), 1245–1253 (1994)

    Article  Google Scholar 

  20. Levy, H., Sidi, M.: Polling systems: application, modeling and optimization. IEEE Trans. Commun. 38, 1750–1760 (1990)

    Article  Google Scholar 

  21. Olsen, T.L., van der Mei, R.D.: Polling systems with periodic server routeing in heavy traffic: distribution of the delay. J. Appl. Probab. 40, 305–326 (2003)

    Article  Google Scholar 

  22. Reiman, M.I.: Some diffusion approximations with state space collapse. In: Modelling and Performance Evaluation Methodology (Paris. 1983), Lecture Notes in Control and Information Sciences, pp. 209–240. Springer, Berlin (1984)

  23. Resing, J.A.C.: Polling systems and multitype branching processes. Queueing Syst. 13, 409–426 (1993)

    Article  Google Scholar 

  24. Srinivasan, M.M.: Nondeterministic polling systems. Manag. Sci. 37, 667–681 (1991)

    Article  Google Scholar 

  25. Takagi, H.: Analysis of Polling Systems. MIT Press, Cambridge, MA (1986)

    Google Scholar 

  26. Titchmarsh, E.C.: Theory of Functions. Oxford University Press, London (1939)

    Google Scholar 

  27. van der Mei, R.D.: Distribution of the delay in polling systems in heavy traffic. Perform. Eval. 38, 133–148 (1999)

    Article  Google Scholar 

  28. van der Mei, R.D.: Towards a unifying theory on branching-type polling systems in heavy traffic. Queueing Syst. 57, 29–46 (2007)

    Article  Google Scholar 

  29. van der Mei, R.D., Winands, E.M.M.: Heavy traffic analysis of polling models by mean value analysis. Perform. Eval. 65, 400–416 (2008)

    Article  Google Scholar 

  30. van Wijk, A.C.C., Adan, I.J.B.F., Boxma, O.J., Wierman, A.: Fairness and efficiency for polling models with the \(\kappa \)-gated service discipline. Perform. Eval. 69, 274–288 (2012)

    Article  Google Scholar 

  31. Vanghi, V., Damnjanovic, A., Vojcic, B.: The Cdma 2000 System for Mobile Communications: 3G Wireless Evolution. Prentice Hall PTR, Englewood Cliffs, NJ (2004)

    Google Scholar 

  32. Vishnevskii, V.M., Semenova, O.M.: Mathematical models to study the polling systems. Autom. Remote Control 67, 173–220 (2006)

    Article  Google Scholar 

  33. Weststrate, J.A.: Analysis and Optimization of Polling Models. PhD thesis, Katholieke Universiteit Brabant (1992)

  34. Weststrate, J.A., van der Mei, R.D.: Waiting times in a two-queue model with exhaustive and Bernoulli service. Zeitschrift für Oper. Res. 40, 289–303 (1994)

    Google Scholar 

Download references

Acknowledgments

The authors wish to thank Marko Boon, Sem Borst and Maria Vlasiou for valuable comments on earlier drafts of the present paper. Funded in the framework of the STAR-project “Multilayered queueing systems” by the Netherlands Organization for Scientific Research (NWO). The research of Onno J. Boxma is performed in the IAP Bestcom project, funded by the Belgian government.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Jan-pieter L. Dorsman.

Appendices

Appendix 1: Proof of Lemma 3.1

Proof

We first focus on the value of \(\left| 1-f_1^{(\infty )}(z_2)\right| = \lim _{j \rightarrow \infty } \left| 1-f_1^{(j)}(z_2)\right| \). For arbitrary \(j>0\), we have for any \(z_2\) in the unit circle that

$$\begin{aligned} \left| 1-f_1^{(j)}(z_2)\right|&= \left| 1-f_1(f_1^{j-1}(z_2))\right| \\&= \left| \int \limits _{t=0}^\infty (1-e^{-\lambda _1(1-\widetilde{K}_{1,2}(f_1^{(j-1)}(z_2)))t}) \hbox {d}\mathbb {P}(P_2 < t)\right| \\&\le \int \limits _{t=0}^\infty \left| 1-e^{-\lambda _1(1-\widetilde{K}_{1,2}(f_1^{(j-1)}(z_2)))t}\right| \hbox {d}\mathbb {P}(P_2 < t), \end{aligned}$$

where the inequality constitutes the triangle inequality. Note that \(\left| 1-e^{-x}\right| \le \left| x\right| \) for any \(x \in \{z\in \mathbb {C}: \mathfrak {R}(z) > 0\}\), so that

$$\begin{aligned} \left| 1-f_1^{(j)}(z_2)\right|&\le \int \limits _{t=0}^\infty \lambda _1 t \left| 1-\widetilde{K}_{1,2}(f_1^{(j-1)}(z_2))\right| \hbox {d}\mathbb {P}(P_2 < t) \nonumber \\&= \lambda _1\mathbb {E}\left[ P_2\right] \left| 1-\widetilde{K}_{1,2}(f_1^{(j-1)}(z_2))\right| \nonumber \\&\le \lambda _1\mathbb {E}\left[ P_2\right] \left| \int \limits _{t=0}^\infty (1-e^{-\lambda _2(1-f_1^{(j-1)}(z_2))t}) \hbox {d}\mathbb {P}(P_1<t)\right| \nonumber \\&\le \lambda _1\mathbb {E}\left[ P_2\right] \lambda _2\mathbb {E}\left[ P_1\right] \left| 1-f_1^{(j-1)}(z_2)\right| . \end{aligned}$$
(32)

Iteration of (32) leads to

$$\begin{aligned} \left| 1-f_1^{(j)}(z_2)\right| \le \left( \lambda _1\mathbb {E}\left[ P_2\right] \lambda _2\mathbb {E}\left[ P_1\right] \right) ^j\left| 1-z_2\right| . \end{aligned}$$
(33)

By (1) we have that \(\mathbb {E}[P_i] = \mathbb {E}[B_i](1-\rho _i)^{-1}\), so that

$$\begin{aligned} \lambda _1\mathbb {E}\left[ P_2\right] \lambda _2\mathbb {E}\left[ P_1\right] = \frac{\rho _1}{1-\rho _2}\frac{\rho _2}{1-\rho _1} < 1. \end{aligned}$$
(34)

The inequality follows since the queues are assumed to be stable, i.e., \( 0 \le \rho < 1\). Therefore, \(\rho _1 = \rho -\rho _2 < 1-\rho _2\), and similarly \(\rho _2 < 1-\rho _1\). A combination of (32) and (34) now leads to

$$\begin{aligned} 0 \le \lim _{j \rightarrow \infty } \left| 1-f_1^{(j)}(z_2)\right| \le \lim _{j \rightarrow \infty } \left( \lambda _1\mathbb {E}\left[ P_2\right] \lambda _2\mathbb {E}\left[ P_1\right] \right) ^j\left| 1-z_2\right| = 0. \end{aligned}$$

Since \(\lim _{j \rightarrow \infty } \left| 1-f_1^{(j)}(z_2)\right| = 0\), we must have that \(f_1^{(\infty )}(z_2) = \lim _{j \rightarrow \infty } f_1^{(j)}(z_2) = 1\).

By similar arguments, it can be shown that \(f_2^{(\infty )}(z_1)=1\) for any \(z_1\) in the unit circle. Finally, it is evident that \(\widetilde{K}_{1,2}(1) = \widetilde{K}_{2,1}(1) = \widetilde{F}_1(1,1) = \widetilde{F}_2(1,1) = 1\). The lemma now follows. \(\square \)

Appendix 2: Proof of Lemma 3.2

Proof

We initially focus on the product \(\prod _{j=0}^\infty a_1(f_1^{(j)}(z_2))\). By the theory of infinite products (see e.g., [26, Chapter 1]), we have that \(\prod _{j=0}^\infty a_1(f_1^{(j)}(z_2))\) converges iff \(\sum _{j=0}^\infty (1-a_1(f_1^{(j)}(z_2)))\) converges. To establish the latter, it is enough to prove that \(\sum _{j=0}^\infty \left| 1-a_1(f_1^{(j)}(z_2))\right| \) converges. We observe that

$$\begin{aligned}&\left| 1-a_1(f_1^{(j)}(z_2))\right| \nonumber \\&\quad = \left| 1-\frac{r_{2,1}\widetilde{M}_{2,1}(\widetilde{K}_{1,2}(f_1^{(j)}(z_2)), f_1^{(j)}(z_2))}{1-r_{1,1}\widetilde{M}_{1,1}(\widetilde{K}_{1,2}(f_1^{(j)}(z_2)), f_1^{(j)}(z_2))} \frac{r_{1,2}\widetilde{M}_{1,2}(\widetilde{K}_{1,2}(f_1^{(j)}(z_2)), f_1^{(j)}(z_2))}{1-r_{2,2}\widetilde{M}_{2,2}(\widetilde{K}_{1,2}(f_1^{(j)}(z_2)), f_1^{(j)}(z_2))}\right| \nonumber \\&\quad = \left| \frac{\sum _{i=1}^2 A_{1,i}(f_1^{(j)}(z_2))(1-\widetilde{M}_{i,1}(\widetilde{K}_{1,2}(f_1^{(j)}(z_2)), f_1^{(j)}(z_2)))}{D(z_2)}\right. \nonumber \\&\qquad + \left. \frac{\sum _{i=1}^2 A_{2,i}(f_1^{(j)}(z_2))(1-\widetilde{M}_{i,2}(\widetilde{K}_{1,2}(f_1^{(j)}(z_2)), f_1^{(j)}(z_2)))}{D(z_2)}\right| , \end{aligned}$$
(35)

where

$$\begin{aligned} A_{1,1}(z_2)&= r_{1,1}(1-r_{2,2}), \\ A_{1,2}(z_2)&= (1-r_{1,1})(1-r_{2,2}), \\ A_{2,1}(z_2)&= (1-r_{1,1})(1-r_{2,2})\widetilde{M}_{1,2}(\widetilde{K}_{1,2}(z_2), z_2),\\ A_{2,2}(z_2)&= r_{2,2}(1-r_{1,1}\widetilde{M}_{1,1}(\widetilde{K}_{1,2}(z_2), z_2)) \hbox { and } \\ D(z_2)&= (1-r_{1,1}\widetilde{M}_{1,1}(\widetilde{K}_{1,2}(f_1^{(j)}(z_2)), f_1^{(j)}(z_2)))\\&\times (1-r_{2,2}\widetilde{M}_{2,2}(\widetilde{K}_{1,2}(f_1^{(j)}(z_2)), f_1^{(j)}(z_2))). \end{aligned}$$

Using the triangle inequality and similar arguments as those in the proof of Lemma 3.1, we note that for \(1 \le i,k \le 2\) and \(j>0\),

$$\begin{aligned}&\left| 1-\right. \left. \widetilde{M}_{i,k}(\widetilde{K}_{1,2}(f_1^{(j)}(z_2)), f_1^{(j)}(z_2))\right| \\&\quad \le \int \limits _{t=0}^\infty \left| 1-e^{-(\lambda _1(1-\widetilde{K}_{1,2}(f_1^{(j)}(z_2)))+\lambda _2(1-f_1^{(j)}(z_2)))t}\right| \hbox {d}\mathbb {P}(S_{i,k} <t) \\&\quad \le \mathbb {E}\left[ S_{i,k}\right] \left( \lambda _1\left| 1-\widetilde{K}_{1,2}(f_1^{(j)}(z_2))\right| + \lambda _2\left| 1-f_1^{(j)}(z_2)\right| \right) \\&\quad \le \mathbb {E}\left[ S_{i,k}\right] \lambda _2(\lambda _1\mathbb {E}\left[ P_1\right] +1)\left| 1-f_1^{(j)}(z_2)\right| . \end{aligned}$$

Moreover, it is trivially seen that \(\left| A_{i,k}(z_2)\right| \le 1\) for \(1 \le i,k \le 2\) and any \(z_2\) in the unit circle. Furthermore, since \(\left| \widetilde{M}_{i,k}(\widetilde{K}_{1,2}(z_2), z_2)\right| \le 1\), we have that \(|D(z_2)| \ge (1-r_{1,1})(1-r_{2,2})\). Therefore, a combination of (33) and (35) with the triangle inequality leads to

$$\begin{aligned} \left| 1-a_1(f_1^{(j)}(z_2))\right|&\le \frac{\mathbb {E}\left[ S_{1,1}\right] +\mathbb {E}\left[ S_{1,2}\right] +\mathbb {E}\left[ S_{2,1}\right] +\mathbb {E}\left[ S_{2,2}\right] }{(1-r_{1,1})(1-r_{2,2})}\\&\times \,\lambda _2(\lambda _1\mathbb {E}\left[ P_1\right] +1)\left( \lambda _1\mathbb {E}\left[ P_2\right] \lambda _2\mathbb {E}\left[ P_1\right] \right) ^j\left| 1-z_2\right| \end{aligned}$$

This result obviously shows, in combination with (34), that \(\sum _{j=0}^\infty \left| 1-a_1(f_1^{(j)}(z_2))\right| \) is bounded from above by a converging geometric sum. As such, \(\sum _{j=0}^\infty \left| 1-a_1(f_1^{(j)}(z_2))\right| \) converges, so that \(\prod _{j=0}^\infty a_1(f_1^{(j)}(z_2))\) converges. The convergence of the product \(\prod _{j=0}^\infty a_2(f_2^{(j)}(z_1))\) can be established similarly. \(\square \)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Dorsman, Jp.L., Boxma, O.J. & van der Mei, R.D. On two-queue Markovian polling systems with exhaustive service. Queueing Syst 78, 287–311 (2014). https://doi.org/10.1007/s11134-014-9413-y

Download citation

  • Received:

  • Revised:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11134-014-9413-y

Keywords

Mathematics Subject Classification

Navigation