Keynote Speakers

Keynote 1 

Causal Software Aging: Models for Estimating, Intervening, and Explaining  

Roberto Pietrantuono 

Abstract 

The software aging community has developed a rich toolbox over the years: statistical trend detection and time-to-exhaustion prediction on the measurement side, analytical models — from Markov chains to stochastic Petri nets — for rejuvenation scheduling on the modeling side. This foundation makes it possible to ask a further class of questions, interventional and counterfactual in nature: what would happen to resource depletion if we acted on a workload knob, a configuration parameter, a suspicious process? Would that failure have occurred had we rejuvenated earlier? After all, rejuvenation itself is, and has always been, a causal intervention. In this talk I introduce structural causal models as a class of models this community can leverage, and show what they enable on aging telemetry: confounder-free estimation of aging trends and time to exhaustion; intervention planning via what-if analysis to select mitigations that causally affect degradation; and counterfactual root-cause explanation of past failures. I will discuss how these models extend predictive notions of causality already explored in aging analysis (Granger, transfer entropy) toward interventional ones, and how they can work in tandem with the Markovian tradition, opening opportunities toward a causal software aging engineering.  

Biography 

Roberto Pietrantuono is an Associate Professor at the University of Naples Federico II, where he works in the Dependable Systems and Software Engineering Research Team (DESSERT). His research interests are in software engineering, particularly software testing, software quality assurance, and software reliability engineering, including a long-standing line of work on software aging and rejuvenation  spanning aging analysis in operating systems, cloud and edge platforms, and stream processing systems. 

His current research focuses on Causal Software Engineering: the use of causal models and causal inference to support quality assurance, root cause analysis, and governance of cmoplex software systems. He has co-authored about 130 papers in these areas and coordinates the EU project uDEVOPS (www.udevops.eu) on microservices and DevOps. He co-founded Critiware (www.critiware.com), a company working in critical systems engineering, and Ai.Res (https://ai-res.it/en), working on responsible AI. He is a Senior Member of the IEEE and a member of the ACM. 

Keynote 2
Toward AI systems aging and rejuvenation    

Fumio Machida 

Abstract 

Software systems are increasingly embodied with AI and ML components. Like hardware and software components, AI models in the system are susceptible to aging, which degrades their performance over long-running operation. This talk begins with lessons learned from a few decades of software aging research, and then discusses a research direction toward AI systems aging and rejuvenation. The central argument is the similarities and differences between software and AI components in aging and rejuvenation mechanisms. Many techniques developed in software reliability research can be leveraged to analyze aging of AI systems and to plan proactive mitigations. On the other hand, several new challenges arise from the special characteristics of AI models' aging, distribution drift, and uncertainty, and maintenance techniques, such as retraining, fine-tuning, and repairing. The talks will cover several recent studies that will pave the way for this research direction and shape the new realm of engineering research in AI systems maintenance.  
 
Biography
Fumio Machida is an associate professor at the Dependable AI Systems (DAIS) Laboratory in the Computer Science Department of the University of Tsukuba. He was a principal researcher at NEC Corporation until 2019 and was a visiting scholar in the Department of Electrical and Computer Engineering at Duke University in 2010. He was a recipient of the Young Scientists’ Prize of Japan in 2014. He was named a distinguished contributor of the IEEE Computer Society, class of 2023. He has contributed to organizing communities for several top conferences, including serving as general chair of the 35th IEEE International Symposium on Software Reliability Engineering. His research interests include modeling and analysis of system dependability, software aging and rejuvenation, and reliability of AI systems. He is a senior member of the IEEE and a member of ACM.