Sharding

Sharding is a scaling technique that splits a blockchain's validation work across multiple parallel subsets of nodes, called shards, so that no single node has to process the entire network's transaction load. Vitalik Buterin frames it as a way to satisfy all three legs of the blockchain scalability trilemma—scalability, decentralization, and security—at once, by randomly sampling validators into committees that each verify a different shard in parallel rather than concentrating trust in a small group or a single chain. The technique that makes sharding safe without forcing every node to download every shard's full data is data availability sampling, where light nodes verify small random pieces of a shard's data instead of the whole thing; see data availability. Sharding is distinct from rollups, which scale by moving execution off-chain rather than by splitting the base chain's own validation work. For tokenized domains and other on-chain assets, a sharded base chain's throughput headroom matters most at the layer 2 level, where most everyday domain transactions actually settle.