Blink: Fast Connectivity Recovery Entirely in the Data Plane

Thomas Holterbach, Edgar Costa Molero, Maria Apostolaki, Alberto Dainotti, Stefano Vissicchio, Laurent Vanbever

UCL Discovery (University College London) · 2019 · 73 citations · 13 references

Concepts

Abstract

We present Blink, a data-driven system that leverages TCPinduced signals to detect failures directly in the data plane.
\nThe key intuition behind Blink is that a TCP flow exhibits a
\npredictable behavior upon disruption: retransmitting the same
\npacket over and over, at epochs exponentially spaced in time.
\nWhen compounded over multiple flows, this behavior creates
\na strong and characteristic failure signal. Blink efficiently
\nanalyzes TCP flows to: (i) select which ones to track; (ii)
\nreliably and quickly detect major traffic disruptions; and (iii)
\nrecover connectivity—all this, completely in the data plane.
\nWe present an implementation of Blink in P4 together with
\nan extensive evaluation on real and synthetic traffic traces.
\nOur results indicate that Blink: (i) achieves sub-second rerouting for large fractions of Internet traffic; and (ii) prevents
\nunnecessary traffic shifts even in the presence of noise. We
\nfurther show the feasibility of Blink by running it on an actual
\nTofino switch.

References

13