Reducing large adaptation spaces in self-adaptive systems using machine learning

Modern software systems often have to cope with uncertain operation conditions, such as changing workloads or fluctuating interference in a wireless network. To ensure that these meet their goals uncertainties be mitigated. One approach realize this is self-adaptation equips system feedback loop. The loop implements four core functions – monitor, analyze, plan, and execute share knowledge the form of runtime models. For large number adaptation options, i.e., spaces, deciding which option select for may time consuming even infeasible within available window make an decision. This particularly case when rigorous analysis techniques are used formal verification at runtime, widely adopted. technique deal options reducing space using machine learning. State art has showed effectiveness technique, yet, systematic solution able handle different types lacking. In paper, we present ML2ASR+, short Machine Learning Adaptation Space Reduction Plus. Central ML2ASR+ configurable learning pipeline supports effective spaces threshold, optimization, setpoint goals. We evaluate two applications sizes spaces: Internet-of-Things application service-based system. results demonstrate can applied reduce hence decisions over 90%, negligible effect on realization