Scalable and Flexible IoT data analytics: when Machine Learning meets SDN and Virtualization

Abstract

This paper deals with Internet of Things (IoT) data analytics in a collaborative platform where computing resources are available both at the network edge and at the backend cloud. Thereby, the requirements of both low-latency and delaytolerant IoT applications can be met. Moreover, this platform faces the challenging heterogeneous features of IoT data, i.e. its high dimensionality or its geo-distributed and streaming data nature. The proposed approach relies on two pillars. On the one hand, recent advances of machine learning (ML) techniques are leveraged to describe how the IoT data analytics can be performed in our platform. On the other hand, the virtualization, centralized management, global view and programmability of the computing and network resources is considered to fulfill the requirements of the ML methods. Unlike the related work, herein the interplay and synergies between those two pillars is explained. Also the ML methods for this collaborative platform are described in more detail.