Network reliability problem
l Random edge failures
» Estimate FAIL(p) = Pr[graph disconnects]
l Naïve Monte Carlo simulation
» Chernoff bound---“tight concentration”
Pr[ |X-E[X]|  m e E[X] ]   <   exp(-e2E[X]/4)
» O(log n / e2FAIL(p)) trials expect O(log n / e2)
network failures---good for Chernoff
» So estimate within e in O*(m/e2FAIL(p)) time