Parameterisation and Optimisation Patterns for MAPE-K Feedback Loops
Verena Kloes, Thomas Goethel, Sabine Glesner · 2017
Self-adaptive systems employ feedback loops to cope with possibly unpredictable environment changes. These feedback loops not only have to capture a suitable adaptation logic, but also have to adhere to non-functional constraints like timing and qualities like precision. Several design options influence these aspects. However, they are often not explicitly accessible in feedback loops and may interfere with each other. Thus, the design is complex and time-consuming. To overcome these problems, we introduce (1) a parameterisation scheme for the widely-known MAPE-K feedback loop to make design options explicit and adjustable, and (2) present goal-oriented optimisation patterns, which capture the interrelations between design options and their effect on non-functional constraints and adaptation qualities. As a result, we enable (self-) optimisation w.r.t. constraints and qualities at design-time and also at run-time. We illustrate the application of our approach using a collective self-adaptive production system.