The BANYAN project organizes a 3-day training school focused on new tools, techniques, models and paradigms that are relevant to the characterization, operation and management of current- and futuregeneration networks and systems. Experts in the field, including from partners of the BANYAN project, will provide a number of lectures on relevant topics. The summer school will also be an occasion for ESRs to present their on-going research activities and obtain feedback from senior attendees. The current document contains the agenda of the event.
The Summer School will be held in the premises of IMDEA Networks Institute, located in Avenida del Mar Mediterraneo 22, 28918 Leganés, Madrid, Spain, from May 24 to May 26, 2023. The program of the school listed bvelow. All times are Central Europe Summer Time (CEST). Titles, abstracts and short bios of all presenters are provided in the remainder of the document.
Jose Ayala Romero
“Energy-Efficient Orchestration of virtualized RANs: Experimental Analysis and Bayesian Learning)”
Virtualized base stations (vBS) can be implemented in diverse commodity platforms and are expectedto bring unprecedented operational flexibility and cost efficiency to the next generation of cellular networks. However, their widespread adoption is hampered by their complex configuration options that affect in a non-traditional fashion both their performance and their power consumption. Following an indepth experimental analysis in a bespoke testbed, we characterize the vBS power consumption profile and reveal previously unknown couplings between their various control knobs. Motivated by these findings, we develop a Bayesian learning framework for the orchestration of vBSs and design algorithms for three different use cases: (i) A vBS in which we want to balance performance and energy consumption, (ii) a battery-powered vBS where there are power constraints, and (iii) a vBS and an AI edge service, where the configuration needs to be done jointly.
Short bio – Jose A. Ayala Romero received his Ph.D. degree from Technical University of Cartagena, Spain, in 2019. Then, he joined Trinity College Dublin, Ireland, as a post-doctoral researcher. From January 2021 to September 2022, he was a Senior Data Scientist at Huawei Ireland Research Center, Ireland. Currently, he is a Senior Researcher at NEC Laboratories Europe, Germany. His research interests include the application of machine learning to resource allocation and energy efficiency problems in mobile networks.
“Three, Two, One, Ignition and Lift-off – What it Takes for Open Source Research Software to Reach the Stars”
In this talk, I will revisit a couple of audacious research challenges we solved in my team. Following a death-or-glory approach, we tackled a number of black-box systems in the area of wireless and mobile networking as well as in the area of cybersecurity and privacy. Our goal was to gain a deep understanding of the system and then improving the feature set, performance, security, privacy, or more. To this end, developing own prototypes and research software was paramount to success. I will spotlight selected results that had profound impact on either the state-of-the-art in research or on solving real-world problems. By discussing the journeys to obtain these results, I will share with you on how to ultimately “reach the stars”.
Short bio – Matthias Hollick is a Full Professor of Computer Science at the Technical University of Darmstadt where he is heading the Secure Mobile Networking Lab (SEEMOO) since 2009. He is coaffiliated with the Department of Electrical Engineering and Information Technology Department. His research has been published at top venues such as ACM CoNEXT, ACM IMWUT, ACM MobiCom, AMC MobiSys, ACM Sigmetrics, ACM WiSec, EWSN, IEEE ComST, IEEE JSAC, IEEE Infocom, IEEE JIoT, IEEE S&P, IEEE TMC, IEEE/ACM ToN, NDSS, PETS, USENIX Security, and other renowned venues in the topic area of mobile and wireless systems as well as cybersecurity and privacy. The research of Prof. Hollick’s team has been awarded with more than 15 best paper and demo awards including at ACM CHI, ACM MobiCom, ACM IMWUT, EWSN, ACM WiSec, and IEEE DOCSS in recent years. The results of his research found their way into numerous highly visible open source software projects. Matthias Hollick is acting as the Speaker and Scientific Coordinator of the emergenCITY Research Center on Digital Resilience established in the Hessian State Excellence Initiative LOEWE in January 2020. He further acts as the Deputy Speaker of the DFG Doctoral School on Privacy and Trust for Mobile Users and as the Co-Deputy Speaker of the DFG Collaborative Research Center 1053 MAKI on Multi-mechanisms Adaptation in the Future Internet. From 2019 to 2022 he was a member of the
Board of ATHENE, the National Research Center for Applied Cybersecurity, which is among the three largest cybersecurity centers in Europe.
“An introduction to Interconnection in the Mobile Ecosystem”
The IP eXchange (IPX) Network interconnects about 800 Mobile Network Operators (MNOs) worldwide and a range of other service providers (such as cloud and content providers) to enable global data roaming. Global roaming now supports the fast growth of the Internet of Things (IoT), while it also responds to the insatiable demand coming from digital nomads, who adhere to a lifestyle where they connect from anywhere in the world. In this talk, we’ll take a first look into this so-far opaque mobile ecosystem, and present first-of-its-kind in-depth analysis of an operational IPX Provider (IPX-P). The IPX Network is a private network formed by a small set of tightly interconnected IPX-Ps. We analyze an operational dataset from a large IPX-P that includes BGP data as well as statistics from signaling. We shed light on the structure of the IPX Network as well as on the temporal, structural and geographic features of the IPX traffic. Our results are a first step to fully understand the global mobile Internet, especially since it now represents a pivotal part in connecting IoT devices and digital nomads all over the world. Finally, we discuss the different operator models, the limitations of current “global” operators on the market that leverage the IPX Network, and how we envision the next generation global operator model.
Short bio – Andra Lutu is a Senior Researcher at Telefónica Research, in Madrid, Spain. Her main research interests lie in the areas of network measurements, interdomain routing and mobile networks, and her work translated into multiple publications in top venues such as ACM MobiCom, ACM SIGCOMM or ACM IMC. As part of Telefónica Research, Andra has been the recipient of an H2020 Marie Curie Individual Fellowship grant funding her work on “Dynamic Interconnections for the Cellular Ecosystem (DICE)”, which is partly included in this talk.
“Primer on Online Convex Optimization”
We will start the presentation with the general Online Learning (OL) procedure. We will present one of the most important algorithms within the Online Cover Optimization (OCO) framework, the Online Mirror Descent (OMD), and we will give a brief intuitive connection to how OMD can adapt to different geometries of the feasible space in order to give solutions with performance guarantees. Then we will present Regret, one of the main metric considered in OCO, and the main OL benchmarks against which online algorithms are compared. We will see some results on the no-regret property (no-regret, i.e. asymptotic optimality, and negative results). We will also see some specific research works on which OCO has been applied as a solution and conclude by wrapping up with some use cases and discuss possible future directions where OCO could be a good tool and those where it is not an adequate tool, focusing on problems that arise within networking.
Short bio – Livia Elena Chatzieleftheriou is a post-doctoral researcher at the IMDEA Networks Insitute and the University of Carlos III in Madrid. She holds an M.Sc. in applied mathematics and she was awarded her Ph.D. in Computer science working on resource allocation, content recommendations, and Online Learning (OL) mechanisms for mobile edge networks after presenting her work in top-tier conferences and journals in her field. Her current research interests are in online learning and explainable AI for next-generation mobile networks, and her research aims at designing innovative approaches to Network Intelligence (NI) and automation for Zero-Touch Networks (ZTNs), and at learning adaptive resource scheduling policies for 6G.
“A data-driven approach to network reconstruction in large-scale dynamical social networks”
Interpersonal influence estimation from empirical data is a central challenge in the study of social structures and dynamics. Opinion dynamics theory is a young interdisciplinary science that studies opinion formation in social networks and has a huge potential in applications, such as marketing, advertisement and recommendations. The term social influence refers to the behavioral change of individuals due to the interactions with others in a social system, e.g. organization, community, or society in general. The advent of the Internet and the explosion of Online Social Networks such as Facebook and Twitter have made a huge volume of data easily available that can be used to measure social influence over large populations. In this short course, we provide a brief overview on modern techniques aimed at qualitatively and quantitatively inferring social influence from data using a systems and control viewpoint. The lecture is structured into four blocks, divided as follows.
- We first introduce some definitions and models of opinions dynamics and review some structural constraints of online social networks, based on the notion of sparsity.
- We review the main approaches to infer the network’s structure from a set of observed data.
- We present some algorithms that exploit the introduced models and structural constraints, focusing on the sample complexity and computational requirements.
- Finally, we present recent applications to the problem of automatically detecting bots in fringe social networks.
Short bio – Fabrizio Dabbene is a Director of Research at the institute IEIIT of the National Research Council of Italy (CNR), where he is the coordinator of the Information & Systems Engineering Group. He has held visiting and research positions at The University of Iowa, at Penn State University and at the Russian Academy of Sciences, Institute of Control Science, Moscow. His research interests include data-driven robust methods for systems and control, design of advanced guidance and navigation schemes for aerospace applications, and analysis of tecno-social dynamical networks. On these topics, he has published more than 100 research papers and two books. He is recipient of the 2010 EurAgeng Outstanding Paper Award. He served as Associate Editor for Automatica (2008-14) and for the IEEE Transactions on Automatic Control (2008-12), and he is currently Senior Editor of the IEEE Control Systems Society Letters. He was elected member of the Board of Governors (2014-16) and he served as Vice-President for Publications (2015-16). He is currently chairing the IEEE-CSS Italy Chapter.
“Characterizing and Modeling Session-Level Mobile Traffic Demands from Large-Scale Measurements”
We analyze transport-layer sessions generated by a wide range of mobile services at over 22,000 cells of an operational mobile network, and carry out a statistical characterization of their associated traffic volume and temporal duration. Our measurement-based study unveils previously unobserved sessionlevel behaviors that are unique to individual mobile applications and persistent across space and time. Based on these insights, we model the distribution of the session-level load and its relationship with the session duration for a variety of services, using simple yet effective mathematical tools. Our models are complementary to existing ones that aim at reproducing either packet-level statistics or traffic demands aggregated over all sessions in space and time. They thus offer an original angle to mobile traffic data generation, and support a more credible performance evaluation of solutions for network planning and management, including a more dependable training and testing of data-driven tools. We assess the utility of the models in a practical application use case, demonstrating how their realism enables a more trustworthy evaluation of energy-efficient allocation of compute resources in virtualized radio access networks.
Short bio – André Zanella is a PhD researcher at IMDEA Networks Institute. He received his bachelor’s and master’s degrees in Telecommunications Engineering at the Federal University of Parana, in Brazil. He is currently a part of the Networks Data Science group, under Dr. Marco Fiore, working on mobile traffic analysis and remote sensing with network metadata, where his interests lay in developing techniques that help solving social sciences problems using information gathered by network operators.
Aristide Tanyi-Jong Akem
“Machine Learning Inference in Programmable Switches with Random Forests”
User-plane machine learning enables low-latency and high-throughput inference at line rate. Yet, data planes are highly constrained environments, and restrictions are especially marked in programmable switches with limited available memory and minimum support for mathematical operations or data types. As a result, current solutions for in-switch inference that are compatible with production-level hardware lack support for complex features or suffer from limited scalability, which creates performance barriers in complex tasks involving large decision spaces. We address this limitation and present two solutions. First, we present Henna, a first in-switch hierarchical inference framework that operates at packet-level and splits large classification tasks into smaller ones that are easier to solve with models that fit switch constraints. Next, we present Flowrest, a first complete Random Forest (RF) model implementation that can operate at the level of individual flows in commercial switches. Flowrest builds on (i) a novel framework to embed generic flow-level machine learning models into programmable switch ASICs, and (ii) original guidelines for tailoring RF models to operations in programmable switches already at the design stage. We implement Henna and Flowrest as open-source software using the P4 language and assess their performance in an experimental platform based on Intel Tofino switches. Tests with tasks of unprecedented complexity show how our models can improve accuracy relative to existing packet-level solutions by up to 21% in the case of Henna and by up to 39% with Flowrest.
Short bio – Akem is a PhD student in the Networks Data Science Group at IMDEA Networks Institute in Madrid, Spain. He is also a student at Universidad Carlos III de Madrid, where he is enrolled in the Telematics Engineering program. Prior to his PhD studies, he completed an engineering degree at the University of Yaounde I, in Cameroon and a master’s in electrical and computer engineering at Carnegie Mellon University Africa, in Rwanda. There, his research was on machine learning for energy-saving optimization in cellular networks. He is currently involved with the European Union’s Horizon 2020 project “BANYAN” which aims to bring big data analytics to radio access networks. At the moment, he is visiting Orange Labs in Paris, France as part of the secondments of the project. Akem’s current research interest is in the area of in-band network intelligence, with a focus on in-network machine learning.
Gabriel O. Ferreira
“Mixed-Integer Optimization Approaches for Power and Resource Allocation in OFDMA Wireless Networks”
A Sequential Mixed-Integer Optimization approach to minimize the base stations’ power consumption in an orthogonal frequency-division multiple access wireless data network (OFDMA) is proposed. In these scenarios, the powers of different base stations cause cross interference and the frequency channel is divided, allowing multiple access. Besides, the assignments between users and base stations and the working bandwidths are also problem variables, leading to a non-convex mixed-integer problem. The proposed relaxation introduces a combination of change-of-variables and linearization, and leads to a sequence of mixed integer linear or quadratic programs, which can be efficiently handled by modern solvers. We also present an extension of the aforementioned method, based on piecewise concave approximation of the non-convex constraints.
Short bio – Gabriel O. Ferreira received the B.S. degree in mechatronics engineering from the Federal Center for Technological Education of Minas Gerais, Brazil, in 2018 and his M.Sc. degree in Electrical Engineering from the Federal University of São João del-Rei, Brazil, in 2020. He is currently pursuing the Ph.D. degree with the Department of Electronics and Telecommunication, Politecnico di Torino, Italy. His research interests include modeling and forecasting mobile network traffic, network optimization, optimization problems, and network closed-loop control.
“Characterizing mobile service demands and optimal indoor radio access network deployment for 5G and beyond systems with data-driven propagation models”
Cell densification through the installation of small cells in indoor environments is an emerging solution to enhance the operation of wireless networks. However, the deployment of new components within the heart of the radio access network calls for expedient tools that can assist and ensure their optimal placement within the existing network infrastructure. This talk will probe how data-driven propagation models can be forged with synthetic data derived from high-performance propagation solvers or realworld measurements, aiming at bridging the efficiency-accuracy gap that the conventional propagation models exhibit. In particular, an expedient and credible data-driven radio propagation model will be presented, called EM DeepRay, which builds on a convolutional encoder–decoder to replicate in a few milliseconds the results of a ray tracer, by encoding physics-based information of an indoor environment
and decoding it as the path loss heatmap. Consequently, the excellence of the model of the model will be demonstrated in practical use cases, such as the wireless channel uncertainty quantification and the optimal network planning. To further highlight the importance of indoor wireless networks and the necessity of robust tools that can guide their deployment, a comprehensive study on the particularities of the traffic generated by the indoor base stations of a major mobile network operator will follow.
Short bio – Stefanos Bakirtzis (Member, IEEE) received the Diploma degree in electrical and computer engineering from the National Technical University of Athens, Athens, Greece, in 2017, and the M.A.Sc. degree in electrical and computer engineering from the University of Toronto, Toronto, ON, Canada, in 2020. He is currently pursuing the Ph.D. degree with the Department of Computer Science and Technology, University of Cambridge, Cambridge, working on the Big Data Analytics for Radio Access Networks (BANYAN) Project as a member of Ranplan Wireless Network Design Ltd., Cambridge. His research interests include wireless communication systems, network optimization, wireless channel modeling, machine learning and artificial intelligence, computational modeling, and stochastic uncertainty quantification. Mr. Bakirtzis received the Onassis Foundation Scholarship and the Marie Skłodowska-Curie Actions-Innovative Training Networks (MSCA-ITN) Fellowship. He has been a finalist in the 2020 IEEE International Microwave Symposium Advanced Practices Paper Competition Award and in the 2023 IEEE International Conference on Acoustics, Speech, and Signal Processing First Pathloss Radio Map Prediction Challenge, whilst he was a member of the team placing third, among the award-winning teams, in the GBSense Challenge 2022.
“Deep learning-assisted indoor & outdoor radio map prediction”
Radio map is essential as network planning entails it identifying signal coverage and signal strength, such that the best location for the base station can be found while minimizing deployment resources. A radio map is essential as network planning entails it to identify signal coverage and signal strength, such that the best location for the base station can be found while minimizing deployment resources. We utilize deep learning-based approaches to guarantee a fast and accurate radio map prediction.
Short bio – Kehai Qiu received his B.Eng. degree from the Beijing University of Posts and Telecommunications in 2015, and his MSc. degree from the University of Sheffield in 2020. He is a Ph.D. candidate at the Computer Laboratory, University of Cambridge, and also works as a Doctoral Research Fellow in Ranplan Wireless, UK.