Schedule: 02/06/2016 – 16:00-17:00
Self-stabilization is a versatile technique for distributed systems and networks to withstand transient faults and attacks of arbitrary nature and extent: after faults and attacks cease to hit the distributed system, it recovers by itself in finite time. This talk surveys a number of recent results related the quantitative evaluation of self-stabilizing distributed systems. One of the key metrics for evaluating the efficiency of self-stabilization is the stabilization time, that is the maximum amount of time one has to wait before normal service is restored after failures. We argue that this metric is not necessarily the best one when it comes to evaluating the actual performance of a self-stabilizing distributed system. In particular, the stabilization time may not be a good indicator when performance metrics are to be considered, including the expected recovery time from failures. We present simulation and probabilistic model checking techniques that permit to get a much finer grained view of the recovery process. This more accurate benchmarking can in turn be used to design performance-centric algorithms or improve their implementation.
Sebastien Tixeuil is a Professor at Pierre and Marie Curie University in Paris since 2007, where he currently leads the Networks and Systems department of LIP6. His research interests include faults and attacks tolerance in networks and distributed systems, communication complexity, passively and actively mobile networks of sensors, and formal methods for distributed computing. He obtained his M.S. in Theoretical Computer Science from Paris-Sud University and his M.S. in Applied Computed Science from Pierre and Marie Curie University, both in 1995. After a PhD obtained in 2000 from Paris-Sud University, he remained there as Associate Professor until 2007. He is a honorary member of Institut Universitaire de France.