Using Adaptive Probing for Real-Time Problem Diagnosis in Distributed Computer Systems

摘要: In this work, we focus on cost-efficient techniques for realtime diagnosis in distributed systems that allow an adaptive, on-line selection and execution of appropriate measurements (tests). Particularly, one our applications concerns fault computer networks by using test transactions, or probes (e.g., "traceroute" "ping" commands). The key efficiency issues include both the cost probing number probes), computational complexity diagnosis. past work (see (Rish, Brodie, & Ma 2002a)), derived some theoretical conditions required asymptotic error-free diagnosis, developed efficient search probe set can greatly reduce size while maintaining its diagnostic capability (Brodie, Rish, 2001). Next, considered problem real-time as a probabilistic inference Bayesian investigated simple local approximation techniques, based variable-elimination (the minibucket scheme (Dechter Rish 2002)). Our empirical studies show these approximations "degrade gracefully" with noise often yield optimal solution when is low enough, initial analysis explains behavior simplest (greedy) 2002a; 2002b). future will adapting more sophisticated such Generalized Belief Propagation (Yedidia, Freeman, Weiss 2001), to scenarios, real-time, incremental learning Dynamic Networks historic data