International Workshop on Teaching Performance Analysis of Computer Systems
| 09:15 | Opening Remarks |
| 09:25 | Talk 1: Dieter Fiems |
| 10:05 | Talk 2: Michela Meo |
| 10:45 | coffee break |
| 11:15 | Q&A for Talk 1 and Talk 2 |
| 11:55 | Discussion Session (D1): Performance Education in a Data-Driven World |
| 12:35 | Talk 3: Ana-Lucia Varbanescu |
| 13:15 | Lunch |
| 14:45 | Talk 4: Cristina L. Abad |
| 15:25 | Q&A for Talk 3 and Talk 4 |
| 16:15 | Coffee Break |
| 16:45 | Discussion Session (D2): Motivating Students |
| 17:25 | Closing Remarks |
Abstract: In the age of machine learning, traditional performance analysis courses face challenges such as decreasing student interest and increasing competition within study programmes. At the same time, courses need to account for increasingly heterogenous groups of students, both in terms of background, interests and mathematical skill. This paper presents a personal view on teaching performance evaluation techniques. We argue that stochastic modelling should take centre stage in a performance analysis course, and that stochastic analysis is a means to an end, rather than the main focus.
Bio: Dieter Fiems obtained an M.Sc. in Electrical Engineering Technology from KAHO-St-Lieven in 1997 and a PhD in Engineering from Ghent University in 2004. He is currently Associate Professor at the Department of Telecommunications and Information Processing at Ghent University. His research interests include various applications of stochastic processes in operations research and performance analysis of communication networks. In particular, he is interested in applications of queueing theory and game theory in wireless and vehicular networks as well as in healthcare and inventory management. His current courses include stochastic simulation, game theory, traffic flow modelling and Bayesian statistics.
Abstract: I am teaching performance evaluation to students in a M.Sc. focused on ICT for Smart Societies. This program aims to equip students with the necessary knowledge and skills in ICT to drive innovation across various engineering fields. The focal point is communication networks, viewed as a collection of enabling technologies fundamental to enhancing engineering through information-driven approaches. Given the multi-disciplinary nature of the program, the students require methodological tools that can help them understand problems in different contexts. Additionally, these tools should encourage them to approach problems from different levels of abstraction. The effort to approach problems from multiple perspective is addressed by using tools, such as analytical modeling and simulation, in a complementary way.
During the talk I would like to share my experience and explore a question that often crosses my mind: Is it worthwhile to teach this?
Bio: Michela Meo is a Professor of Telecommunication Engineering at Politecnico di Torino, Italy. Her research interests include green networking, energy-efficient mobile networks and data centers, Internet traffic classification and characterization, and machine learning for video quality of experience. From 2015 to 2022, she chaired the International Advisory Council of the International Teletraffic Conference. She has contributed to several special issues of international journals. She serves as a Senior Editor of IEEE Transactions on Green Communications and has previously held positions as Associate Editor for ACM/IEEE Transactions on Networking, the Green Series of the IEEE Journal on Selected Areas of Communications, and IEEE Communication Surveys and Tutorials. In various capacities, she has overseen the organization of numerous conferences such as MSWiM, ACM e-Energy, ITC, Infocom Miniconference, ICC symposia, and ISCC.
Abstract: With computing systems and applications growing increasingly complex and heterogeneous, modern performance engineering must combine different methods and tools, offering different trade-offs between user-intervention, accuracy, and automation. These methods and tools rely on performance models and methods ranging from analytical (white-box) models and careful hand-tune algorithmic optimizations to data-driven/AI-driven (black-box) models and autotuning.
In this talk, I argue these methods and tools must form a toolbox that any performance engineering student and professional should have. Moreover, this approach improves the accessibility of the trade, as it allows users to select the tools most suitable for their skills.
I will further argue that performance engineering education could resemble an apprenticeship model, where we teach our students how to make best use of these tools, while we encourage them to practice by creatively solving real-life performance problems. In my experience, this approach leads not only to creative solutions for performance engineering, but also significant improvements to the toolbox instead. I will make these points using examples from our research and education in performance engineering, and I will conclude with a few thoughts on how to embed more performance engineering elements in regular CS, EE, and AI curricula.
Bio: Ana Lucia Varbanescu holds a BSc and MSc degree from POLITEHNICA University in Bucharest, Romania, and a PhD in Computer Science from TUDelft, The Netherlands. Since 2022, she is also Professor at the University of Twente. She has been a visiting researcher at IBM TJ Watson (2006, 2007), Barcelona Supercomputing Center (2007), NVIDIA (2009), and Imperial College of London (2013).
She is one of the founding members of the CompSys community in the Netherlands. In recent years, she was program co-chair for several conferences, including ISC, HPDC, HiPC, CCGrid, and ICS.
Ana’s research stems from HPC, and investigates the use of heterogeneous systems for high-performance applications, with a special focus on performance and energy efficiency modeling for both scientific and data-intensive applications. Her latest research focuses on zero-waste computing and model-based systems co-design.
Ana has been teaching for more than 20 years in BSc, MSc, and graduate programs in different parts of Europe, from Romania to Spain. She has created several HPC courses (for different audiences and systems). She is an advocate of combining theory and practice in education, and a pioneer of performance engineering education in The Netherlands, where she created and taught the first course on the topic.
Abstract: In CS curriculum guidelines, Performance Engineering has been typically listed as a small, elective component, if at all. Even less has been said about how and when to teach workload modeling. In this talk, I will talk about how Systems courses are a good place to include this topic, including suggestions on how to do so that will be rooted in personal experience and existing literature. Tying to the TeaPACS discussion, workload modeling is inherently data-driven, and as such, the inclusion of this topic can help students get motivated in Performance Engineering, as many of them are already working on acquiring data science skills that they can naturally use for workload modeling.
Bio: Cristina L. Abad is a Professor in the Department of Electrical Engineering and Computer Science at Escuela Superior Politécnica del Litoral in Guayaquil-Ecuador, where she leads the Distributed Systems Research Lab and co-directs the Big Data Research Group. She received her Ph.D. in 2014 from the University of Illinois at Urbana-Champaign. For three years during her PhD, she was a Software Engineering Intern in the Hadoop Core Team at Yahoo, where she worked on workload modeling and evaluation of the HDFS and had the opportunity to contribute to the Apache Hadoop codebase. Her research interests lie at the intersection of Distributed Systems and Performance Engineering. In particular, she works on designing and building distributed systems that can self-adapt to workload changes and maximize performance, with applications in cloud computing and Big Data. Her international funding sources have included VLIR-UOS, Google, Microsoft, Amazon Web Services, and AT&T Labs Research. She has received a Fulbright Fellowship, a UIUC CS Excellence Fellowship, and two Google Faculty Research Awards. Cristina is a member of IEEE, ACM, SPEC RG, and Usenix, and is the elected Secretary of the SPEC RG.
