ECQT2025

PROGRAMME

Plenary Talks

Joris Walraevens (Ghent University, Belgium) 

Title: Novel Multiclass Queueing Models for Road Traffic Intersections

Abstract: In this talk, we explore novel multiclass models that are motivated by road traffic. Multiple vehicle types typically use the same roads simultaneously. These different types do not only differ in attainable speeds but also in feasible accelerations and decelerations. For instance, freight vehicles accelerate at a slower pace than normal vehicles leading to slower queue dissipation at intersections and creating gaps in these queues. We model this by a two-class queueing model (e.g.\ normal vehicles and freight vehicles) in which the service rate of the queue depends on the presence of one of the classes (slower rate when freight vehicles are present). We discuss different alternative models depending on the instants the service rate switches. Typical performance measures to be studied are (a) amount of occupied road (expressed in vehicle equivalents) and (b) wasted time, but the distribution of the length of gaps that origin from different accelerations of the different vehicle types could be of interest too.

This is based on joined work with Sara Sasaninejad and Sabine Wittevrongel.

Bionote: Joris Walraevens received the M.S. degree in Electrical Engineering and the Ph.D. degree in Engineering in 1997 and 2004 respectively, all from Ghent University, Belgium. In September 1997, he joined the SMACS Research Group, Department of Telecommunications and Information Processing, at the same university. He was appointed fullt-ime associate professor at Ghent University in 2012 and has been a full professor from 2022 onwards. Since October 2018, he is also the head of Department of Telecommunications and Information Processing. His main research interest throughout his career are multi-class models and their applications in diverse domains such as telecommunication networks, healthcare and transportation networks.

Tuan Phung-Duc  (University of Tsukuba, Japan)

Title: Queueing models for autoscaling systems: Exact and asymptotic solutions
Abstract: In many modern computer systems such as data centers or 5G systems, computing resources are scaled in/out to adapt with varying demands and to save energy consumption. However, scaling in/out the resources may incur some setup delay, which potentially increases the response time and energy consumption. In this study, we present the queue analysis of such a system to quantify the tradeoff between delay performance and energy consumption, in which we derive efficient algorithms for the stationary distribution of the underlying multi-dimensional queueing processes. We then present the analysis of the response time distribution, which is further utilized for a game-theoretical analysis of such models with strategic customers. We also extend to some models with abandonments or retrials for which scaling limits are also considered.

Bionote: Tuan Phung-Duc is an Associate Professor at Institute of Systems and Information Engineering, University of Tsukuba. He received a Ph.D. in Informatics from Kyoto University in 2011. He is on the Editorial Board of the Journal of the Operations Research Society of Japan and Queueing Models and Service Management. He served as Guest Editor of four special issues of Annals of Operations Research. He was the Chairman of the 10th International Workshop on Retrial Queues (WRQ'2014) and the TPC co-chair of the 23rd International Conference on Analytical, and Stochastic Modelling Techniques and Applications (ASMTA'16), TPC co-chair of The 13th and 14th International Conference on Queueing Theory and Network Applications (QTNA2018, QTNA2019), General co-chair of EAI VALUETOOLS 2020 - 13th EAI International Conference on Performance Evaluation Methodologies and Tools, and General chair of ASMTA/EPEW: The 26th International Conference on Analytical and Stochastic Modelling Techniques and Applications / 17th European Performance Engineering Workshop. Dr. Phung-Duc is an Elected Member of IFIP WG 7.3 (Working Group on Computer System Modeling). He received the Research Encourage Award from The Operations Research Society of Japan in 2013 and several Best Paper Awards in International Conferences. In 2023, he was selected as a top 2% researcher in the world (by Stanford University based on Scopus). His research interests include Applied Probability and Stochastic Modelling of Computer and Service Systems.

 Maria Vlasiou (University of Twente, The Netherlands)

Title: Large fork-join queuing networks 

Abstract: Fork-join queues are a model central to parallel computing systems and high-tech manufacturing. In such systems, arriving jobs are parallelised in subtasks over many servers or suppliers providing components. The performance of these systems is determined by the slowest subtask in the system. Fork-join queues with $N$ parallel servers capture these dynamics, but present mathematical challenges due to the common arrival process. In this presentation, we leverage various limiting regimes (fluid limits, diffusion limits, large deviations and extreme-value theory) for the performance analysis of $N$-server fork-join networks, with $N$ large, and apply these insights to optimise fork-join systems.

Specifically, for nearly deterministic arrival and service times, we present a fluid limit for the maximum queue length as $N\to\infty$. Remarkably, this fluid limit depends on the initial number of tasks. We refine these results by examining the tail asymptotics for the delay in a Brownian fork-join network, where we see that the rare-event probability of a large backlog behaves roughly as a power law with $N$, where three regimes can be distinguished: one that exhibits a form of asymptotic independence, one with a highly irregular behaviour that has a clear dependence structure among the $N$ suprema, and the third being a nontrivial transition on the boundary. For general but independent arrival and service distributions, we prove that the scaled maximum steady-state waiting time and the scaled maximum steady-state queue length converge to a normally distributed random variable as $N\to\infty$. We provide further limiting theorems for heavy-tailed service systems, where we provide the first known sharp convergence result for the maximum of a fork-join queue with heavy-tailed characteristics. Last, we sketch how to leverage these results in high-tech manufacturing to achieve the joint optimisation of the component production capacity and inventory, leading to asymptotically optimal methods to dimension the system. To derive these results, we develop extreme value theory and diffusion approximations for fork-join systems, combining them with literature on heavy-tails and methods from optimisation.

Bionote: Maria Vlasiou is a Professor at the University of Twente, the Netherlands, holding a Hypatia chair, a Research Fellow of the European research institute EURANDOM, and the Director of the Dutch Network on the Mathematics of Operations research LNMB. In addition, she represents the Netherlands at the Committee for Women in Mathematics of the International Mathematical Union, is the country coordinator for the Netherlands at the European Women in Mathematics association, and is on the advisory boards of the women in mathematics networks of Greece and of Cyprus.

Her research centres on the performance of stochastic processing interacting networks. Her research strives to make networked systems sustainable and resilient, focusing on their complex interactions. She develops new mathematical tools in performance evaluation, optimisation, control, and decision making and applies these tools to design new algorithms that can be deployed in data centres, the electricity grid, communication systems, high-tech manufacturing and beyond. Her research has been funded by grants exceeding 37M from more than 10 science foundations, universities, societies, and organisations.

Prof. Vlasiou has received the best paper award in ICORES 2013, the Marcel Neuts student paper award in MAM8, the 3rd prize of the 8th conference in Actuarial Science, and the 2022 the UPS George D. Smith prize of INFORMS. She has chaired the IFIP Performance 2023 international symposium on Computer Performance, Modeling, Measurements and Evaluation and the 58th Dutch National Mathematics Congress. She currently serves as an Associate Editor for the journals IISE Transactions, Mathematical Methods of Operations Research, OR Spectrum, and INFOR, and Guest Editor for Performance Evaluation.

 Mark Squillante (IBM Thomas J. Watson Research Center, USA)

Title:

Abstract:

Bionote: Mark S. Squillante is a Distinguished Research Scientist and Manager in the Mathematical Sciences Department of IBM Research at

the Thomas J. Watson Research Center.  He received his Ph.D. degree from the University of Washington.  His research interests broadly concern

the mathematical foundations of the analysis, modeling and optimization of the design, control and performance of stochastic systems.  He has

published over 300 papers and 70 issued/filed patents/disclosures, and has received The Best Publication in Applied Probability Award (INFORMS

APS), The Daniel H. Wagner Prize (INFORMS), 9 best paper awards, 19 keynote/plenary presentations, and 27 major IBM technical awards. He

currently serves as Editor-in-Chief of Stochastic Models, as Chair of IFIP Working Group 7.3, and on the Board of Directors of INFORMS and

AACC. He is an elected Fellow of INFORMS, SIAM, ACM, IEEE, and AAIA, and member of AAAS, AMS, Bernoulli Society, and IMS.