Politecnico di Torino | Servizi per la didattica

KEYWORD

Inference of network behavior with low-resource machine learning

azienda Thesis in external company estero Thesis abroad

keywords EDGE COMPUTING, MACHINE LEARNING, NETWORKING

External reference persons Leonardo Linguaglossa, Telecom Paris, linguaglossa@telecom-paris.fr

Research Groups Telecommunication Networks Group

Thesis type EXPERIMENTAL - DEVELOPMENT, RESEARCH

Description Recent years witnessed a trend of "softwarization" of network components. Instead of static, expensive hardware, operators have started to adopt a more flexible approach based on Virtual Network Functions. This paradigm (aka Network Function Virtualization) advocates implementing network middleboxes such as firewalls or NATs as pieces of software to be deployed and executed on commercial off-the-shelf (COTS) hardware. Similarly, in 5G networks, the traditional radio access is moving to cloud RAN or virtualized RAN running on general purpose computing and radio components. This has boosted the development of several packet processing frameworks and software switches, which show nowadays multi 10-Gbps capabilities in COTS servers. In parallel, network systems are increasingly adopting machine learning (ML) techniques to solve complex networking tasks such as traffic classification or resource allocation. While there is a natural question about the feasibility of embedding such ML components into todays' networks elements (i.e., routers or base stations), two main challenges arise in the context of high-speed software networks.
First, ML techniques require a large amount of data to be collected for both training and validation: when done in software, measurements can highly affect the measured values, thus biasing the collected data. The intensity of this becomes stronger when measurements are taken close to the data path. Second, even after the training phase, complex model calculations may require dedicated hardware such as external GPUs or custom hardware designed for neural network processing such as TPUs or VPUs. In this project, we develop a novel approach based on non-invasive data collection and the integration of compact ML techniques in standard COTS equipment. Relying on pure software, our methodology allows the integration of advanced management functionalities in existing network infrastructure with no additional equipment. The candidate will work on (i) low-impact network measurements with both direct and indirect observations; (ii) inference/predictive modeling of a complete system with ML and/or classical approaches; (iii) deployment of low-resource models for runtime query/action operations and automated recovery. The project (acronym: IONOS-DX) has received an individual grant from the ANR (French Agency of Research).