Analysis of MAPE-K Loop in Self-adaptive Systems for Cloud, IoT and CPS

Abstract

Self-adaptive approaches aim to address the complexity of modern computing generated by the runtime variabilities and uncertainties. In this context, MAPE-K loop is considered today a major approach for the design and implementation of self-adaptive solutions because it captures in a systematic way the main steps of the adaptation process: (1) Monitor the execution context, (2) Analyze the monitored context, (3) Plan the appropriate adaptation strategy, (4) Execute the adaptation strategy, all these steps using a common Knowledge about the context. Implementations of MAPE-K loops may be particularly complex, domain specific, as well as case study dependent. In this paper, we provide a preliminary analysis of MAPE-K loops in various artifacts in different application domains (i.e., cloud - Hogna and TMA, cyber-physical systems - TRAPP and AMELIA, Internet of Things - DeltaIoT). Our main objective is to outline the similarities and differences among the available implementations of MAPE-K control feedback loops in self-adaptive systems. Additionally, the application domains of the considered examples are highly related, so that solutions in one domain may trigger developments in others. We also provide an insight into MAPE-K loops to enable researchers and practitioners to use, re-use, improve the available solutions.