Paper Review: Gossip-Based Computation of Aggregate Information
Jack W. Rae · 2011
Kempe et al. [2003] claim ‘gossip’ protocols are scalable and robust, becomming increasingly relevant as we encounter network infrastructures that are very complex or unstable. The paper introduces several ‘gossip’ protocols to compute aggregate information such as sums, averages and random samples of ‘nodes’ within a network. A metric of ‘information diffusion’ is introduced, which they then use to prove the claimed exponential rate of convergence.