Self-Adaptation Based on Big Data Analytics: A Model Problem and Tool

摘要

In this paper, we focus on self-adaptation in large-scale software-intensive distributed systems. The main problem in making such systems self-adaptive is that their adaptation needs to consider the current situation in the whole system. However, developing a complete and accurate model of such systems at design time is very challenging. To address this, we present a novel approach where the system model consists only of the essential input and output parameters. Furthermore, Big Data analytics is used to guide self-adaptation based on a continuous stream of operational data. We provide a concrete model problem and a reference implementation of it that can be used as a case study for evaluating different self-adaptation techniques pertinent to complex large-scale distributed systems. We also provide an extensible tool for endorsing an arbitrary system with self-adaptation based on analysis of operational data coming from the system. To illustrate the tool, we apply it on the model problem.