In this talk, I will present FlightPath, a novel peer-to-peer
streaming application that provides a highly reliable data stream to a
dynamic set of peers. FlightPath offers a more stable stream than
previous works by several orders of magnitude. I will explain the
techniques we use to maintain such stability despite peers that act
maliciously and selfishly.
More broadly, this talk will discuss the core of FlightPath's success:
approximate equilibria. I will highlight how these equilibria let us
rigorously design incentives to limit selfish behavior, yet also
provide the flexibility to build practical systems. Specifically, I
will show how we use epsilon-Nash equilibria to engineer a live
streaming system to use bandwidth efficiently, absorb flash crowds,
adapt to sudden peer departures, handle churn, and tolerate malicious
activity.