Probe scheduling for efficient detection of silent failures》.pdf

Performance Evaluation 79 (2014) 73–89 Contents lists available at ScienceDirect Performance Evaluation journal homepage: /locate/peva Probe scheduling for efficient detection of silent failures Edith Cohen a,b,∗, Avinatan Hassidim c,d , Haim Kaplan b , Yishay Mansour b , Danny Raz e,c , Yoav Tzur c a Microsoft Research, CA, USA b Blavatnik School of Computer Science, Tel Aviv University, Israel c Google, Inc. Israel RD Center, Israel d Bar-Ilan University, Israel e Technion, Israel a r t i c l e i n f o a b s t r a c t Article history: Most discovery systems for silent failures work in two phases: a continuous monitoring Available online 11 July 2014 phase that detects presence of failures through probe packets and a localization phase that pinpoints the faulty element(s). We focus on the monitoring phase, where the goal is to Keywords: balance the probing overhead with the cost associated with longer failure detection times. Silent failures We formulate a general model for the underlying fundamental subset-test scheduling Active probing Scheduling problem. We unify the treatment of schedulers and cost objectives and make several con- tributions: We propose Memoryless schedules—a natural subclass of stochastic schedules which is simple and suitable for distributed deployment. We show that the optimal mem-