How can complex AI-based microservice chains be configured efficiently?
2025/07/01

In dynamic and heterogeneous edge environments, the deployment of AI-driven mission-critical applications is becoming increasingly complex. Each application must fulfill strict constraints in terms of performance, resource availability, and timing. Traditional configuration methods often fail under such conditions.
In this context, we are pleased to announce that researchers from TK will present their paper titled “FM4MC: Improving Feature Models for Microservice Chains – Towards More Efficient Configuration and Validation” at the prestigious International Conference on Software Engineering (ICSE 2026).
The FM4MC project introduces a novel methodology to efficiently configure AI-based microservice chains. At its core is an innovative decomposition of large feature models into Partial Feature Models (PFMs). These are preprocessed during an offline phase, enabling significant reductions in both storage and computational overhead. In the subsequent online phase, FM4MC can derive valid and executable configurations in real time, tailored to the current edge hardware environment, even in mission-critical, resource-constrained scenarios.
The evaluation results are impressive: FM4MC enables a 1,000× speedup during runtime configuration and reduces storage consumption by up to 100×. These advances allow for real-time decision-making in mission-critical systems, whether in autonomous driving, edge-based surveillance, or emergency response applications.
We warmly congratulate the research team on this achievement and look forward to the broader impact of FM4MC in the field of adaptive AI software systems.
Citation information:
Gropengießer, U., Wolfart, P., Liphardt, J., Mühlhäuser, M. (2026). FM4MC: Improving Feature Models for Microservice Chains—Towards More Efficient Configuration and Validation [to appear].
