Data Management Platform for Smart Orchestration of Decentralized and Heterogeneous Vehicular Edge Networks

Abstract

Intelligent edge orchestration has become a vital component within next generation communication networks, such as 5G. They offer optimal resource allocation and service distribution, hence allow for full utilization of the opportunities provided by those networks. Orchestrators make use of Machine Learning (ML) techniques to determine the most optimal operational decisions, such as deployment and scaling of services, ensuring the quality of the service performance. The training and validation of these models require significant amount of data. However, in such environments we deal with heterogeneous and distributed data sources, in which the data needs to be collected and pre-processed efficiently, and as such, made ready-to-use for these ML models. Hence, in this paper several state-of-the-art data management technologies suitable for edge computing and orchestration are investigated and compared. After investigating the theoretical features of these technologies, they are deployed and tested on the Smart Highway testbed. The strengths and shortcomings of these systems are presented and compared based on the Quality of Service (QoS) requirements for various vehicular services dealing with highly mobile users, i.e., vehicles, which are considered as quite stringent. This hybrid platform, combining several data management technologies, will be used to assist and enrich the research on smart edge orchestration at IDLab and serve as a reference for other interested parties.