"Consensus" algorithms are decision proceedures by which a group of individual elements (e.g. processors, robots, sensors, maybe even people), perhaps each with a distinct an different initial position, come to "agree" on a single common position.  "Gossip" algorithms are distributed asynchronous proceedures in which pairs of individuals in a network each successively and temporarily move toward a dyadic consensus.  Under very general conditions we are able to prove exponential convergence to global consensus, and establish estimates on the rate of convergence based on network structure and rates of pairwise convergence.