A Customizable Architecture for Application-Centric Management of Context-Aware Applications

Context-aware applications present common requirements (e.g., heterogeneity, scalability, adaptability, availability) in a variety of domains (e.g., healthcare, natural disaster prevention, smart factories). Besides, they do also present domain specific requirements, among which the application concept itself is included. Therefore, a platform in charge of managing their execution must be generic enough to cover common requirements, but it must also be adaptable enough to consider the domain aspects to meet the demands at application-level. Several approaches in the literature tackle some of these demands, but not all of them, and without considering the applications concept and the customization demands in different domains. This work proposes a generic and customizable management architecture that covers both types of requirements based on multi-agent technology and model-driven development. Multi-agent technology is used to enable the distributed intelligence needed to address many common requirements, whereas model-driven development allows to address domain specific particularities. On top of that, a customization methodology to develop specific platforms from this generic architecture is also presented. This methodology is assessed by means of a case study in the domain of eHealthCare. Finally, the performance of MAS-RECON is compared with the most popular tool for the orchestration of containerized applications.