Performance Modelling in the Age of Data and AI

Abstract

 Performance Modelling is concerned with understanding and predicting the dynamic behaviour of systems (typically technological systems), particularly from the perspective of fair and efficient use of resources. Originally based on Queueing Theory, as technology became more sophisticated a variety of different modelling paradigms were developed and adopted over the last forty years. 

 The increasing prominence of AI systems has brought a number of challenges in terms of the systems to be modelled and mechanisms for modelling them. In this talk I present some work combining mechanistic performance modelling techniques with more data driven approaches. I will also look at some of the challenges of defining performance for systems such as foundation AI models where resource consumption is rarely taken into account

Our Speaker

 Jane Hillston is Professor of Quantitative Modelling in the School of Informatics at the University of Edinburgh. She received a BA in Mathematics from the University of York (UK) in 1985, a MSc in Mathematics from Lehigh University (USA) in 1987 and a PhD in Computer Science from the University of Edinburgh in 1994. She has held a number of leadership roles within the University of Edinburgh, including in Head of School of Informatics from 2018—2023 and Deputy Vice Principal Research 2020—2022. Since 2024 she is Dean of Research Culture and REF in the College of Science and Engineering. 

 Her principal research interests are in formal approaches to modelling the dynamic behaviour of discrete event systems, particularly from the perspective of performance expressed as logical properties. This has included everything from cloud computing, to biological processes, to transport systems in smart cities. Her research has been supported by numerous grants from public funding bodies and industry. Her current research focus is on the performance of AI systems and the humans who interact with them.