[행사] [2023.11.24(Fri.)]AI Convergence Network & Artificial Intelligence Colloquium
- 소프트웨어융합대학교학팀
- 김성민
- Create Date 2023-11-17
- Views 229
![▶](https://fonts.gstatic.com/s/e/notoemoji/15.0/25b6/32.png)
![▶](https://fonts.gstatic.com/s/e/notoemoji/15.0/25b6/32.png)
![▶](https://fonts.gstatic.com/s/e/notoemoji/15.0/25b6/32.png)
![▶](https://fonts.gstatic.com/s/e/notoemoji/15.0/25b6/32.png)
Abstract : Measurements in high-speed networks are very important, yet very challenging to obtain. As any analysis is only as good as the obtained observations, if measurements do not correspond to the system under test, this will impact all the subsequent analysis. In particular, when networking is implemented in software - as is the case for Network function virtualization (NFV) - measurement operations can highly affect the state of the system being measured. In this presentation we tackle the problems of data acquisition and analysis in NFV networks and present a novel approach based on indirect, non-invasive data collection and the integration of compact ML techniques in standard commercial off-the-shelf equipment. Our objective is to provide a methodology for advanced operation, administration, and maintenance (OAM) functionalities that can be easily integrated into the existing network infrastructure. This presentation is organized as follows: (i) we present the context, positioning and objectives of our project; (ii) we present our proposed methodology and (iii) we showcase our results for one use-case, namely the inference of network state using indirect measurements.
![▶](https://fonts.gstatic.com/s/e/notoemoji/15.0/25b6/32.png)
![▶](https://fonts.gstatic.com/s/e/notoemoji/15.0/25b6/32.png)