From s.vissicchio at ucl.ac.uk  Mon Jul 16 17:02:53 2018
From: s.vissicchio at ucl.ac.uk (Vissicchio, Stefano)
Date: Thu Jul 19 12:09:40 2018
Subject: [Nets-seminars] Talk by Thomas Holterbach (ETH)
Message-ID: <68BD6808-B24A-4249-B364-72CAAB68955D@ucl.ac.uk>

Hello,

Thomas Holterbach, a PhD student at ETH Zurich, will be giving a talk on Friday, about our ongoing collaboration.
Location, time and details of the talk follow.

All invited to attend, feedback on the work would be more than welcome!

Best,
Stefano

----

When: Friday 20th July, 5.30 pm

Where: Room 405 in 66-72 Gower Street

Title: Fast Connectivity Recovery Entirely in the Data Plane

Abstract: A mission-critical, unsolved problem in the operation of WANs is to quickly recover connectivity upon Internet traffic disruptions.  Rerouting traffic to a different path use control-plane information (e.g., BGP messages), which is either inherently slow to propagate over the Internet, or completely absent (as for link congestion).  The most effective fast rerouting technologies rely on hardware-based data plane signals (e.g., loss of light in fibers) induced by local failures and are therefore applicable to local failures only.
In this presentation, we will explore new possibilities, created by programmable switches, for fast rerouting upon signals triggered by Internet traffic disruptions. We will describe Blink, a new data-driven system exploiting TCP-induced signals. The key intuition behind Blink is that a TCP flow exhibits a predictable behavior upon disruption: retransmitting the same packet over and over, at epochs exponentially spaced in time. When compounded over multiple flows, this behavior creates a strong and characteristic failure signal. Blink efficiently analyzes TCP flows, at line rate, to: (i) select which flows to track; (ii) detect major traffic disruptions; and (iii) recover data-plane connectivity, using paths compatible with the operator's policies.
We have implemented Blink in P4 and Python, and we have evaluated it on real and synthetic traces. Our initial results indicate that Blink has the potential to often retrieve connectivity within less than 1 seconds, for realistic Internet traffic.

Bio: Thomas Holterbach is a third year PhD student from ETH Zurich. He is doing research in the Networked Systems Group leaded by Prof. Laurent Vanbever. Before joining ETH, he earned a Bachelor and a Master degree in Computer Science from the University of Strasbourg, France. Thomas?s research is about making the Internet routing more efficient, reliable and secure. More particularly, he is trying to improve the routing convergence upon Internet outages. He has strong interests in BGP, Internet measurements, Software-defined Networking and programmable network devices. Thomas is currently visiting UCL until the end of July 2018. At UCL, he is supervised by Dr. Stefano Vissicchio.