Recent perspectives on network programmability and network verification

About

One of the main motivations behind self-driving cars is the promise of making our roads safer. Human errors indeed explain more than 90% of the car accidents nowadays. This situation bears a lot of similarities with computer networks. Today, most network operators still manage their networks manually and human errors, not equipment failures, explain the vast majority of network downtimes. These downtimes can (and often do) have devastating consequences such as:

preventing millions of people from accessing the Internet for hours;

grounding entire airline fleets; or

preventing emergency calls from being placed.

In this talk, I will describe our recent research results in network verification and network analysis and how they can assist network operators in ensuring network correctness and understanding complex network infrastructures. I will first speak about our works on configuration synthesis which aims at automatically generating low-level network configurations out of high-level requirements (or "intents"). I will then speak about our works on "network captioning" and "specifications mining" which enable to automatically extract high-level insights out of low-level networking data (e.g. Netflow statistics, raw router configurations).

About the speaker

Laurent Vanbever is a tenure-track assistant professor at ETH Zurich where he leads the Networked Systems Group. Prior to ETH, he was a Postdoctoral Research Associate at Princeton where he collaborated with Jennifer Rexford. He obtained his PhD degree in Computer Science from the University of Louvain in 2012. His research focuses on making large network infrastructures more manageable, scalable and, secure. Laurent has won several awards for his research including multiple best paper awards and four applied networking research prizes. .

0:00:00 Slide 1

0:02:57 Slide 2

0:03:25 Slide 3

0:03:51 Slide 4

0:04:04 Slide 5