Understanding the impact of architectural decisions is indeed not trivial, even more so when considering non-functional properties, such as software performance, that are affected by external factors. For instance, the system response time may change due to multiple sources of uncertainty, such as heterogeneous workloads and fluctuations in the usage of available resources. This talk presents some methodologies to quantify software performance characteristics of different architectural decisions, so as to predict issues and avoid expensive repairs. To this end, (anti)patterns represent one viable source of knowledge to relate architectural elements and system performance, which allows to identify design decisions leading to poor performance. Recent results on emerging domains such as architecting federated learning systems will be presented, along with challenges and future research directions.

I am Associate Professor in Computer Science at the Gran Sasso Science Institute, Italy. My main research interests include the quantitative modelling and analysis of interacting heterogeneous distributed systems. More details at: http://cs.gssi.it/catia.trubiani