International Conference on Software Architecture (ICSA) 2026
- Post by: admin
- June 22, 2026
- No Comment
Two research works were presented at the International Conference on Software Architecture (ICSA 2026).
Green Autoscaler for Performance Aware Microservices: a Machine Learning Approach
Thanh-Phuc Tran , Abhinandan Roul , Ishara Galbokka Hewage , Mahira Joytu , Roberta Capuano, Eoan O’Dea, Rafiullah Omar, Hergys Rexha, Sebastien Lanfond , Henry Muccini
The first contribution, “Green Autoscaler for Performance Aware Microservices: a Machine Learning Approach,” addresses the environmental impact of cloud-native systems. Traditional solutions such as the Kubernetes Horizontal Pod Autoscaler do not consider the carbon intensity of electricity, potentially leading to unnecessary carbon emissions. The proposed Carbon-Aware Autoscaling (CAAS) system leverages Spatio-Temporal Graph Convolutional Networks (STGCN) to model workload dynamics and inter-service dependencies. By integrating real-time carbon intensity information, the approach demonstrates the potential to reduce carbon emissions by up to 24% while maintaining high system performance.
LLMs for Architectural Refactoring: An Exploratory study on Monoliths to Microservices
Aneesh Sambu, Roberta Capuano, Eoan O’Dea, Karthik Vaidhyanathan, Henry Muccini
The second contribution, “LLMs for Architectural Refactoring: An Exploratory Study on Monoliths to Microservices,” investigates how Large Language Models (LLMs) can support software architects during the migration from monolithic systems to microservices. To assist with the challenging task of identifying appropriate service boundaries, the authors developed an LLM-driven pipeline that generates architecturally meaningful decompositions. The exploratory study shows that LLMs are particularly effective at producing architecturally sound designs, outperforming traditional approaches in terms of Cyclic Independence (+22.7%). Rather than replacing human architects, LLMs are presented as valuable design partners capable of providing technically coherent alternatives for improving system modularity and dependency management.
Together, these works highlight how artificial intelligence and graph-based learning techniques can contribute to the development of more sustainable, modular, and intelligent software architectures.
https://conf.researchr.org/home/icsa-2026