Fully connected mesh network

A fully connected mesh network is one in which every node has a direct connection to every other node in the network, as demonstrated in the following diagram:

If this diagram looks a bit crowded to you, you've already noticed the single biggest drawback of a fully-connected mesh network. It's nearly impossible to scale beyond a certain point, because each new node on the network requires a connection for each previously connected node on the network. The math works out to a quadratic increase in connections for each new node to be added. Moving past a few nodes on the network becomes physically impossible very quickly.

The incredibly high cost of a fully connected mesh network, however, brings with it the most stable and resilient topology possible. No node has to be responsible for packet-forwarding or request switching, because there should be no context in which two nodes are communicating with each other indirectly. Any one node or connection between nodes can go down, and every other node on the network has full connectivity with no loss of performance. A single weak connection between two nodes has zero impact on the performance of any other two nodes. As topologies go, a fully connected mesh is bulletproof. It's also often prohibitively expensive, and so not common in anything but the smallest and most trivial contexts.