Queueing Systems in a Random Environment

Queueing networks with product-form steady-state distribution have found many fields of applications, e.g. production systems, telecommunications, and computer system modeling. The success of this class of models and its relatives stems from the simple structure of the steady-state distribution which provides access to easy performance evaluation procedures. Starting from the work of Jackson [Jac57] various generalizations have been developed. In real world queueing systems are not isolated and interact with their environment. Adding a random environment to a model usually makes the model more realistic but also more complex to analyze. Nevertheless, under some conditions it is still possible to obtain analytical results. A branch of research which recently has found interest are queueing networks in a random environment with product form steady-state distributions. The main theoretical contributions of this thesis are twofold: (i) We develope a general theory that comprise models with stationary product-form distribution in inventory theory in [Sch04] and Jackson networks with unreliable nodes with stationary product-form distribution in [Sau06]. An important property of the resulting general model is that the queueing system and the environment interact in both directions: the queues can influence the environment and the environment can influences the queues. (ii) With respect to applications we show how different models known from literature can be interpreted in terms of the general theory, construct new models in various applications, and develope an approximation method. In Part I we analyze single-queue systems. In Section 1 we introduce a loss system. In Section 2 we generalize product form lost-sales inventory models from [Sch04] and several other published papers with related models as a loss system with exponential service time. The term loss means that customers get lost when the environment stays in some special states -- the blocking states. In Section 2.1.4 we develop an approximation method for system without loss of customers based on loss systems. In Section 2.2 we apply our loss system results in fields different from inventory management: we analyze in detail an unreliable server with preventive maintenance in Section 2.2.4, a node of a wireless sensor network in Section 2.2.5, and a crusher station in open-pit mining in Section 2.2.6. In Section 3 we analyze the Markov chain embedded at departure instants of the loss system. The embedded Markov chains are an important tool for analyzing queueing system with general service times -- the M/G/1/infinity queues. The famous and frequently used result in classical M/G/1/infinity theory is that the steady-state distribution of an M/G/1/infinity system as continuous time process and as embedded Markov chain, observed at departure times, are the same. We show that this is in general not true for the steady-state distribution of loss systems. We use an embedded Markov chain analysis to extend our results from Section 2 to some loss systems with general service times. In Part II we extend our results for a single-queue loss system to Jackson networks in a random environment. We replace the concept of loss of customers by special rerouting regimes. We establish a connection between these rerouting regimes and randomized random walks. In Section 8 we consider systems where the interaction between environment and queuing system depend on the number of customers in the system. This extension finally allows us to include results about Jackson networks with unreliable nodes from [Sau06] as special cases. Warteschlangennetze, deren stationare Verteilung eine Produktform hat, sind in unterschiedlichen Bereichen angewendet worden, zum Beispiel: Produktionssysteme, Telekommunikation und Modellierung von Rechnersystemen. Ihren Erfolg haben sie ihrer einfachen Struktur der stationarer Verteilung zu verdanken. Sie vereinfacht die Analyse von Leistungskenngrosen der zu modellierenden Systeme. Aufbauend auf der grundlegenden Arbeit von Jackson [Jac57] wurden weitere Verallgemeinerungen entwickelt. In der realen Welt sind Warteschlangensysteme nie isoliert. Sie befinden sich in einer Umgebung, mit der sie interagieren. Das Hinzufugen einer Umgebung zu einem reinen Warteschlangenmodell fuhrt haufig zu einem besseren Gesamtmodell. Gleichzeitig wird die mathematische Analyse dieses Modells schwieriger und komplexer. Unter speziellen Bedingungen ist es dennoch moglich, fur derartige Systeme analytische Losungen zu erhalten. Viele solche Systeme, gehoren zu Warteschlangennetzwerken in einer zufalligen Umgebung, deren stationare Verteilung eine Produktform hat. Die Hauptbeitrage dieser Dissertation liegen in zwei Bereichen: (i) Wir entwickeln eine allgemeine Theorie, die Modelle mit stationarer Produktformverteilung aus den Lagerhaltungsmodellen aus [Sch04] und Jackson-Netzwerke mit unzuverlassigen Knoten mit stationarer Produktformverteilung aus [Sau06] gleichzeitig umfast. Eine wichtige Eigenschaft dieses allgemeinen Modells ist, dass die Warteschlangensysteme und die Umgebung sich gegenseitig beeinflussen: die Warteschlangen konnen die Umgebung beeinflussen und die Umgebung die Warteschlangen. (ii) Auf Anwendungen bezogen, zeigen wir, dass viele aus der Literatur bereits bekannte Modelle sich mit Hilfe dieser allgemeinen Theorie darstellen lassen. Wir stellen neue Modelle fur unterschiedliche Bereiche vor und entwickeln ein Naherungsverfahren. In Teil I untersuchen wir Systeme mit einer einzigen Warteschlange. In Abschnitt 2 stellen wir ein Verlustsystem (loss system) vor . In Abschnitt 2verallgemeinern wir Lagerhaltungsmodelle mit Kundenverlust aus [Sch04] und mehrere ahnliche publizierte Modelle. Diese Verallgemeinerung bezeichnen wir als Verlustsystem. Der Bergriff Verlust (loss) bezieht sich auf die Modellannahme, dass Kunden verlorengehen, solange die Umgebung in sogenannten blockierenden Zustanden ist. In Abschnitt 2.1.4 entwickeln wir ein Naherungsverfahren fur Systeme ohne Kundenverlust, das auf Systemen mit Kundenverlust basiert. In Abschnitt 2.2 nutzen wir Ergebnisse fur Verlustsysteme in anderen Bereichen auserhalb der Lagerhaltung: wir untersuchen detailliert ein Warteschlangensystem mit einem unzuverlassigen Bediener mit praventiver Wartung in Abschnitt 2.2.4, einen Knoten in einem drahtlosen Sensornetzwerk in Abschnitt 2.2.5 und eine Zerkleinerungsanlage im Bergbau in Abschnitt 2.2.6. In Abschnitt 3 untersuchen wir eingebettete, zu Kundenabgangszeiten beobachtete, Markov-Ketten von Verlustsystemen. Eingebetteten Markov-Ketten sind ein wichtiges Werkzeug fur die Untersuchung der Warteschlangensysteme mit allgemeinen Bedienzeitverteilungen, das heist Warteschlangensysteme vom Typ M/G/1/Unendlich. Ein bekanntes und haufig benutztes Ergebnis in der klassischen M/G/1/Unendlich Theorie ist, dass fur ein M/G/1/Unendlich System die stationaren Verteilungen des Prozesses in stetiger Zeit und dijenige der eingebetteten Markov-Kette, betrachtet zu Kundenabgangszeiten, ubereinstimmen. Wir zeigen, dass dies fur die stationaren Verteilungen eines Verlustsystems im Allgemeinen nicht gilt. Wir benutzen Markov-Ketten-Methoden um unsere Ergebnisse fur exponentielle Bedienzeiten aus Abschnitt 2 in einigen Fallen auf Systeme mit allgemeiner Bedienzeit zu erweitern. In Teil II erweitern wir unsere Ergebnisse fur Verlustsysteme mit einer Warteschlange zu Jackson-Netzwerken in zufalliger Umgebung. Wir ersetzen das Konzept des Kundenverlustes durch spezielle Reroutingregeln. Wir stellen einen Zusammenhang her zwischen unterschiedlichen Reroutingregeln und randomisierten Irrfahrten. Zum Schluss, in Abschnitt 8, erlauben wir zusatzlich, dass die Wechselwirkung zwischen der Umgebung und dem Warteschlangennetz von der Kundenzahl im Gesamtsystem abhangen kann. Diese Erweiterung ermoglicht es, die Ergebnisse uber Jackson-Netzwerke mit unzuverlassigen Knoten aus [Sau06] als Spezialfalle der gemeinsamen Theorie zu erfassen.

[1]  Carl M. Harris,et al.  Fundamentals of queueing theory , 1975 .

[2]  H. Kesten Random processes in random environments , 1980, Advances in Applied Probability.

[3]  A. Lambert Branching Processes: Variation, Growth and Extinction of Populations , 2006 .

[4]  P. Naor,et al.  Some Queuing Problems with the Service Station Subject to Breakdown , 1963 .

[5]  P. Moran,et al.  Reversibility and Stochastic Networks , 1980 .

[6]  Jewgeni H. Dshalalow,et al.  Queueing systems with state dependent parameters , 1998 .

[7]  Antonis Economou,et al.  A CHARACTERIZATION OF PRODUCT-FORM STATIONARY DISTRIBUTIONS FOR QUEUEING SYSTEMS IN RANDOM ENVIRONMENT , 2003 .

[8]  Hans Daduna,et al.  Modeling and Performance Analysis of a Node in Fault Tolerant Wireless Sensor Networks , 2014, MMB/DFT.

[9]  R. Manikandan,et al.  A revisit to queueing-inventory system with positive service time , 2015, Ann. Oper. Res..

[10]  Rasoul Haji,et al.  A queueing system with inventory and mixed exponentially distributed lead times , 2011 .

[11]  V. Ramaswami,et al.  An operator analytic approach to product-farm networks , 1996 .

[12]  Nico M. van Dijk,et al.  On Jackson's product form with 'jump-over' blocking , 1988 .

[13]  W. Hackbusch Iterative Lösung großer schwachbesetzter Gleichungssysteme , 1991 .

[14]  Peter Whittle Scheduling and Characterization Problems for Stochastic Networks , 1985 .

[15]  Jr. Shaler Stidham Optimal Design of Queueing Systems , 2009 .

[16]  Hao Gu,et al.  Performance analysis of a wireless sensor network , 2006, IEEE Wireless Communications and Networking Conference, 2006. WCNC 2006..

[17]  Hans Daduna,et al.  Loss systems in a random environment: steady state analysis , 2015, Queueing Syst. Theory Appl..

[18]  R. Cogburn,et al.  Markov Chains in Random Environments: The Case of Markovian Environments , 1980 .

[19]  B. Conolly Structured Stochastic Matrices of M/G/1 Type and Their Applications , 1991 .

[20]  Chuang Lin,et al.  An Analytical Model for Evaluating IEEE 802.15.4 CSMA/CA Protocol in Low-Rate Wireless Application , 2007, 21st International Conference on Advanced Information Networking and Applications Workshops (AINAW'07).

[21]  Peter G. Taylor,et al.  Calculating the equilibrium distribution in level dependent quasi-birth-and-death processes , 1995 .

[22]  L. Christie,et al.  Queuing with Preemptive Priorities or with Breakdown , 1958 .

[23]  Sauer Cornelia,et al.  Availability Formulas and Performance Measures for Separable Degradable Networks , 2003 .

[24]  Wei Wayne Li,et al.  Several Characteristics of Active/Sleep Model in Wireless Sensor Networks , 2011, 2011 4th IFIP International Conference on New Technologies, Mobility and Security.

[25]  Jianming Liu,et al.  A framework for performance modeling of wireless sensor networks , 2005, IEEE International Conference on Communications, 2005. ICC 2005. 2005.

[26]  Peter W. Glynn,et al.  Stochastic Simulation: Algorithms and Analysis , 2007 .

[27]  Richard J. Gibbens,et al.  Dynamic Routing in Fully Connected Networks , 1990 .

[28]  J. R. Jackson Networks of Waiting Lines , 1957 .

[29]  Leonard Kleinrock,et al.  Queueing Systems: Volume I-Theory , 1975 .

[30]  T. Liggett Interacting Particle Systems , 1985 .

[31]  Srinivas R. Chakravarthy,et al.  Queues with interruptions: a survey , 2014 .

[32]  Hans Daduna,et al.  Stochastic networks with product form equilibrium , 2001 .

[33]  Hans Daduna,et al.  A queueing theoretical proof of increasing property of Polya frequency functions , 1996 .

[34]  Antonis Economou,et al.  Product form stationary distributions for queueing networks with blocking and rerouting , 1998, Queueing Syst. Theory Appl..

[35]  Harry G. Perros,et al.  Closed queueing networks with blocking , 1986 .

[36]  Genji Yamazaki,et al.  Decomposability in queues with background states , 1995, Queueing Syst. Theory Appl..

[37]  Vaidyanathan Ramaswami,et al.  Introduction to Matrix Analytic Methods in Stochastic Modeling , 1999, ASA-SIAM Series on Statistics and Applied Mathematics.

[38]  Jean C. Walrand,et al.  A Distributed CSMA Algorithm for Throughput and Utility Maximization in Wireless Networks , 2010, IEEE/ACM Transactions on Networking.

[39]  K. Waldmann,et al.  Optimal control of arrivals to multiserver queues in a random environment , 1984 .

[40]  Simonetta Balsamo,et al.  Separable solutions for Markov processes in random environments , 2013, Eur. J. Oper. Res..

[41]  Maike Schwarz,et al.  Queueing systems with inventory management with random lead times and with backordering , 2006, Math. Methods Oper. Res..

[42]  B. Avi-Itzhak,et al.  A Many-Server Queue with Service Interruptions , 1968, Oper. Res..

[43]  Nico M. van Dijk,et al.  On Practical Product Form Characterizations , 2011 .

[44]  J. Keilson Markov Chain Models--Rarity And Exponentiality , 1979 .

[45]  Hans Daduna,et al.  Separable networks with unreliable servers , 2003 .

[46]  Achyutha Krishnamoorthy,et al.  A survey on inventory models with positive service time , 2011 .

[47]  Ruslan Krenzler,et al.  Loss systems in a random environment-embedded Markov chains analysis , 2013 .

[48]  Seung Ki Moon,et al.  The M/M/1 queue with a production-inventory system and lost sales , 2014, Appl. Math. Comput..

[49]  K. Mani Chandy,et al.  Open, Closed, and Mixed Networks of Queues with Different Classes of Customers , 1975, JACM.

[50]  Bianca Schroeder,et al.  IGP link weight assignment for transient link failures , 2003 .

[51]  Richard J. Boucherie A Characterization of Independence for Competing Markov Chains with Applications to Stochastic Petri Nets , 1994, IEEE Trans. Software Eng..

[52]  G. Falin,et al.  A heterogeneous blocking system in a random environment , 1996, Journal of Applied Probability.

[53]  Uri Yechiali A Queuing-Type Birth-and-Death Process Defined on a Continuous-Time Markov Chain , 1973, Oper. Res..

[54]  Dhanya Shajin,et al.  Analysis of a Multiserver Queueing-Inventory System , 2015, Adv. Oper. Res..

[55]  Hongyi Wu,et al.  A survey on analytic studies of Delay-Tolerant Mobile Sensor Networks , 2007, Wirel. Commun. Mob. Comput..

[56]  Vidyadhar G. Kulkarni,et al.  Production-inventory systems in stochastic environment and stochastic lead times , 2012, Queueing Syst. Theory Appl..

[57]  Rasoul Haji,et al.  The M/M/1 queue with inventory, lost sale, and general lead times , 2013, Queueing Syst. Theory Appl..

[58]  Hongyi Wu,et al.  Analytic, Simulation, and Empirical Evaluation of Delay/Fault-Tolerant Mobile Sensor Networks , 2007, IEEE Transactions on Wireless Communications.

[59]  John Odentrantz,et al.  Markov Chains: Gibbs Fields, Monte Carlo Simulation, and Queues , 2000, Technometrics.

[60]  Hans Daduna,et al.  Loss systems in a random environment , 2013, 1312.0539.

[61]  Yixin Zhu,et al.  Markovian queueing networks in a random environment , 1994, Oper. Res. Lett..

[62]  S. Stidham,et al.  Individual versus Social Optimization in the Allocation of Customers to Alternative Servers , 1983 .

[63]  F. Kelly,et al.  Networks of queues , 1976, Advances in Applied Probability.

[64]  P BuzenJeffrey Computational algorithms for closed queueing networks with exponential servers , 1973 .

[65]  A. Krishnamoorthy,et al.  Stochastic decomposition in production inventory with service time , 2013, Eur. J. Oper. Res..

[66]  Wei Li,et al.  An Energy-Based Stochastic Model for Wireless Sensor Networks , 2011, Wirel. Sens. Netw..

[67]  Marcel F. Neuts,et al.  Matrix-geometric solutions in stochastic models - an algorithmic approach , 1982 .

[68]  W. Hackbusch Iterative Solution of Large Sparse Systems of Equations , 1993 .

[69]  Simonetta Balsamo,et al.  Analysis of Queueing Networks with Blocking , 2010 .

[70]  Dieter Baum,et al.  A Product Theorem for Markov Chains with Application to PF-Queueing Networks , 2002, Ann. Oper. Res..

[71]  Feng Xia,et al.  A packet buffer evaluation method exploiting queueing theory for wireless sensor networks , 2011, Comput. Sci. Inf. Syst..

[72]  Michele Garetto,et al.  An Analytical Model for Wireless Sensor Networks with Sleeping Nodes , 2006, IEEE Transactions on Mobile Computing.

[73]  Ruslan K. Krenzler,et al.  Jackson networks in nonautonomous random environments , 2016, Advances in Applied Probability.

[74]  Yves Dallery,et al.  Manufacturing flow line systems: a review of models and analytical results , 1992, Queueing Syst. Theory Appl..

[75]  Harry G. Perros Approximation algorithms for open queueing networks with blocking , 1989 .

[76]  A. Economou Stationary Distributions of Discrete-Time Markov Chains in Random Environment: Exact Computations and Bounds , 2004 .

[77]  Devavrat Shah,et al.  Randomized Scheduling Algorithm for Queueing Networks , 2009, ArXiv.

[78]  John F. Meyer,et al.  Performability Modeling of Distributed Real-Time Systems , 1983, Computer Performance and Reliability.

[79]  Maike Schwarz,et al.  M/M/1 Queueing systems with inventory , 2006, Queueing Syst. Theory Appl..

[80]  O. Berman,et al.  Stochastic models for inventory management at service facilities , 1999 .

[81]  L. Tierney A note on Metropolis-Hastings kernels for general state spaces , 1998 .

[82]  Nico M. van Dijk Queueing networks and product forms - a systems approach , 1993, Wiley-Interscience series in systems and optimization.

[83]  Antonis Economou,et al.  Generalized product-form stationary distributions for Markov chains in random environments with queueing applications , 2005, Advances in Applied Probability.

[84]  Ruslan K. Krenzler,et al.  Randomization for Markov chains with applications to networks in a random environment , 2014, 1407.8378.

[85]  Richard J. Boucherie,et al.  Queueing networks : a fundamental approach , 2011 .

[86]  R. Serfozo Introduction to Stochastic Networks , 1999 .