This simulation demonstrates how a piece of information, such as a node failure notification or a data update, spreads exponentially across a simulated cluster as nodes randomly gossip with a few peers each round, converging on full cluster-wide knowledge in a number of rounds proportional to the logarithm of the cluster size.
Set the total number of nodes and the fan-out, meaning how many random peers each node contacts per round, then start the simulation and watch as an initial informed node spreads its update outward round by round. Try disabling or removing some nodes mid-run to see how the gossip still reaches the rest of the cluster despite the failures.
Controls let you adjust the total number of nodes, the fan-out per gossip round, the round interval, and simulate node failures, while the display tracks how many nodes are informed after each round and highlights active gossip exchanges between randomly paired peers.
Did you know that in a cluster of one million nodes, gossip with a modest fan-out can still spread an update to essentially every node in around twenty rounds, because doubling the informed population each round means the required rounds grow only with the logarithm of the cluster size, not the size itself.
This simulation demonstrates how a piece of information, such as a node failure notification or a data update, spreads exponentially across a simulated cluster as nodes randomly gossip with a few peers each round, converging on full cluster-wide knowledge in a number of rounds proportional to the logarithm of the cluster size.
This simulation demonstrates how a piece of information, such as a node failure notification or a data update, spreads exponentially across a simulated cluster as nodes randomly gossip with a few peers each round, converging on full cluster-wide knowledge in a number of rounds proportional to the logarithm of the cluster size.
Set the total number of nodes and the fan-out, meaning how many random peers each node contacts per round, then start the simulation and watch as an initial informed node spreads its update outward round by round. Try disabling or removing some nodes mid-run to see how the gossip still reaches the rest of the cluster despite the failures.
Did you know that in a cluster of one million nodes, gossip with a modest fan-out can still spread an update to essentially every node in around twenty rounds, because doubling the informed population each round means the required rounds grow only with the logarithm of the cluster size, not the size itself.