Dynamic deployment prediction and configuration in hybrid cloud / edge computing environments using influence-based learning

Cloud computing has become the norm when it comes to web applications deployment over the last years. Yet, data-intensive applications have shown a great growth were more and more are being developed on a daily basis. Moreover, edge devices have gained a great boost in performance, allowing them to manage and analyze data very efficiently, thus making edge computing an equally as effective alternative. Taking into consideration the always increasing number of information that is being generated requires an enormous amount of resources, hybrid cloud / edge infrastructures seem to be most effective and promising. However, the resource management in such heterogeneous environments regards a difficult task. For this reason, in this paper, we propose a novel dynamic deployment and resources configuration framework for hybrid cloud / edge computing environments using static code analysis. The proposed framework utilizes an influence-based learning technique, for the prediction of the optimal deployment configurations of applications. The proposed solution is evaluated using a data set generated from 389,000 public repositories from GitHub, which proves that our solution manages to successfully identify the optimal deployment configuration of the applications achieving accuracy of over 90%.