queueing theory

queueing theory
Name	Queueing theory
Focus	Operations research
Employed by	Telecommunications, Computer science, Transportation

queueing theory Queueing theory studies the mathematical analysis of waiting lines and service systems, addressing how items, customers, or data are delayed, processed, and depart from service facilities. It combines probabilistic models, stochastic processes, and optimization techniques to predict performance, allocate resources, and design systems across telecommunications, computing, and transportation industries. Developed through contributions from mathematicians, engineers, and economists, the field informs the design of networks, call centers, and manufacturing plants.

History

Origins of queueing theory trace to work by Agner Krarup Erlang on telephone traffic at Københavns Telefonkompagni and later connections with researchers like David George Kendall and A. K. Erlang's contemporaries in early 20th-century Copenhagen. Developments progressed through interactions with scholars at institutions such as Bell Labs, Columbia University, and Princeton University, and through applications in wartime logistics during World War II and postwar industrial expansion. Key foundational advances involved contributions from John von Neumann and Norbert Wiener in stochastic processes, while later expansions incorporated methods from Leonard Kleinrock and Andrew S. Tanenbaum influenced by the rise of packet-switched networks and computer networking research at Massachusetts Institute of Technology and Stanford University.

Fundamentals and Terminology

Core constructs originate from probability theory and stochastic processes developed by figures at University of Cambridge and University of Oxford, employing terms such as arrival process, service process, queue discipline, and system capacity. Notation conventions, including the widely used Kendall notation, stem from the work of David G. Kendall and have been adopted across textbooks and courses at Cornell University and University of California, Berkeley. System components—servers, buffers, customers, and routers—are analyzed using tools introduced by researchers affiliated with Harvard University and University of Pennsylvania, and linked to performance objectives studied at MIT and Imperial College London.

Mathematical Models and Classifications

Models are classified by arrival and service distributions, number of servers, and queue capacity; archetypal models include birth–death processes and Markovian systems informed by research at Bell Labs and IBM. The M/M/1, M/G/1, and G/G/1 families trace to analytical work by Agner Krarup Erlang and formalizations by David George Kendall; multidisciplinary extensions incorporate heavy-tailed behavior studied by researchers at Carnegie Mellon University and University of Toronto. Networked models such as Jackson networks and Gordon–Newell networks built on concepts developed at Princeton University and University of Illinois Urbana-Champaign describe interactions in interconnected service systems. Advanced classifications involve priority queues, bulk-service models, and retrial queues developed in collaboration with scholars at University of Amsterdam and Ecole Polytechnique Fédérale de Lausanne.

Performance Metrics and Analysis Methods

Typical metrics include average waiting time, queue length, throughput, and blocking probability, derived using techniques from Kolmogorov-style forward equations and renewal theory advanced at Moscow State University and University of Chicago. Little's law, a cornerstone result attributed to work at Bell Labs and popularized by practitioners at AT&T, links average population, arrival rate, and time in system. Analytical methods leverage transform techniques, generating functions, and matrix-analytic methods contributed by researchers associated with Siemens and Nokia research labs. Asymptotic analysis and large-deviations theory, influenced by work at Courant Institute and University of California, San Diego, provide approximations for rare-event probabilities and system tails.

Applications and Case Studies

Applications span telecommunications networks designed by teams at Lucent Technologies and Nokia; data centers and cloud computing architectures developed at Google and Amazon Web Services; and transportation hubs analyzed by planners at Port Authority of New York and New Jersey and Transport for London. Manufacturing and supply-chain examples draw on techniques used at General Electric and Toyota, while healthcare operations research implemented at Mayo Clinic and Johns Hopkins Hospital uses queueing-based scheduling models. Case studies in call-center staffing and airline check-in systems often reference benchmarks and implementations by DHL and FedEx.

Simulation and Computational Methods

When analytic solutions are intractable, discrete-event simulation and Monte Carlo methods—popularized by computational groups at Los Alamos National Laboratory and Sandia National Laboratories—provide numerical estimates. Software ecosystems originating from projects at National Institute of Standards and Technology and IBM Research include simulation libraries, queueing solvers, and performance evaluation tools. Algorithmic advances in exact numerical inversion, matrix-geometric methods, and stochastic Petri nets were advanced by researchers at University of Bologna and Technical University of Munich to handle large-scale models.

Recent Developments and Open Problems

Recent work integrates machine learning approaches from Google Research and DeepMind with traditional queueing models to adapt scheduling policies in real time, while research on heavy-traffic limits and mean-field approximations continues at Courant Institute and Princeton University. Open problems include designing robust control policies under adversarial arrivals studied at Microsoft Research and extending analysis for cloud-native microservices explored at Facebook. Fundamental theoretical challenges involve non-Markovian networks, parameter uncertainty, and multi-class interactions pursued by groups at ETH Zurich and University of Cambridge.

Category:Operations research