LIQUID_EDGE

Liquid edge computing based on distributed machine learning and millimeter-wave radio access

PROGETTI DI RICERCA DI RILEVANTE INTERESSE NAZIONALE (PRIN) – Bando 2017

Schedule

January 2020 - January 2024

Coordination

Prof. Sergio Barbarossa (Università degli Studi di ROMA “La Sapienza”)

Partners

• Università degli Studi di ROMA “La Sapienza”
• Università degli Studi di ROMA Tor Vergata
• Università degli Studi di CASSINO
• Università degli Studi di CATANIA
• Università degli Studi di PERUGIA
• Università degli Studi di PARMA

Abstract

The goal of the LIQUID_EDGE project is to provide mobile devices with delay-sensitive services, or with low latency requirements, through

• (a) a strategy of efficient dynamic orchestration of computing, communication and storage resources;
• (b) the use of innovative information-centric network architectures;
• (c) the use of transmission channels based on millimeter wave technology (mmWave).

The basic idea is to bring processing, communication and network resources to such a level of granularity as to offer the user continuous services with seamless features, delivering them as if they were pervasive liquid. This is achieved through a dense deployment of millimeter wave radio access points, associated with the use of the edge computing network paradigm, and artificial intelligence techniques in the orchestration of stateless microservices and virtualized functions in a unikernel environment. The project is based on the dynamic and synergistic interaction of the following enabling factors:

• (i) millimeter wave wireless interfaces with multiple channel access (multi-link) and multiple access technologies (multi-RAT), in cell-free systems (i.e. detached from the typical cellular architecture of current mobile radio systems);
• (ii) the development of microservices such as to fragment cloud services into many small functions that can be quickly distributed, downloaded and transferred thanks to their small size;
• (iii) the joint optimization of computing, communication and storage resources, achieved dynamically and proactively thanks to machine learning techniques

Personnel involved

• Paolo Banelli (local coordinator)
• Paolo Valigi
• Luca Rugini
• Giuseppe Baruffa