Network Routing

💡 The Plain-English Definition

Network routing is how a Lightning payment finds its path from sender to recipient through a web of payment channels — hopping through one or more intermediate nodes to reach its destination even when the two parties have no direct channel between them.

⤢ Enlarge

🤔 But Why Though?

If Lightning needed a direct channel (a funded two-party connection) between every pair of people who wanted to transact, it would take billions of channels and billions of on-chain transactions to set up. Instead, Lightning routes payments through existing channels — a payment from Alice to Dave can travel Alice → Bob → Carol → Dave if those channels exist, even though Alice and Dave have no direct connection.

Finding that route isn’t simple. Each node keeps its own local copy of the channel graph (the public map of Lightning’s channels and their capacities), but it doesn’t know how the balance inside each channel is split — only the total capacity. A channel with 1 BTC of capacity might be 0.99 on one side and 0.01 on the other, making it effectively useless in one direction, and nothing in the graph reveals that.

Routing algorithms have improved a lot to cope. Early Lightning used source routing (the sender worked out the full path). Modern software uses techniques like probing (sending trial payments to learn a channel’s actual balance), probabilistic pathfinding (estimating likely balances and adjusting after each success or failure), and multipath payments (splitting a payment across several routes when no single one can carry the whole amount).

The reliability of Lightning routing has improved dramatically since the early years. Payments succeed at much higher rates than in 2019, thanks to better pathfinding, a denser channel graph with more well-funded nodes, and AMP (Atomic Multipath Payments — splitting a single payment across several simultaneous paths) for larger amounts.

🌍 The Real-World Analogy

Think of Lightning routing like a postal system with no direct route between every pair of cities. A letter from a small village to a remote town travels through a series of sorting centres — each one only knows to pass it to the next centre, not the whole journey. The routing algorithm decides the sequence of handoffs. If a sorting centre is overwhelmed (a channel’s balance is depleted), it finds another route. The letter arrives without the sender knowing, or caring, which sorting centres handled it.