FRADIS: A Machine Learning-based Multipath Solution for Differentiated Services in a Network Slicing-enhanced Delivery Environment

The increase in number of devices per person driven by the latest wearable and IoT devices poses challenges for network support. Despite advancements like IEEE 802.11ax and 5G New Radio, no single technology or provider can handle the growing data surge and diverse service demands. To meet this demand, a programmable and flexible network infrastructure is essential, supporting various technologies and adopting a software-based architecture with open interfaces. A key concepts in achieving this is Network Slicing (NetSli). In this context, the FRAmework for performance-aware Differentiated Innovative Services (FRADIS) was designed as a comprehensive framework for the 5G and beyond heterogeneous network environment, aiming to facilitate the differentiated delivery of services with diverse requirements. FRADIS integrates machine learning to optimize service-specific approaches, choosing between infrastructure-dependent (traffic engineering) and protocol-based solutions. The framework targets a wide range of services, including smart city monitoring, e-health information, emergency messages, infotainment, targeted advertisements, IoT and sensor data, road traffic navigation, agriculture monitoring, and touristic virtual reality. This paper describes the architecture of FRADIS, aiming to achieve traffic control through NetSli at lower network layers, employing a dynamic traffic characteristics-oriented protocol at the transport layer, and using machine learning for adaptive content delivery at the application layer. Preliminary results indicate the benefits of the proposed framework and its flexibility to support multiple types of rich-media applications.