sharding

Sharding is a way to scale a blockchain by splitting it into several parallel pieces, called shards, each handling its own slice of transactions and data, so the network as a whole can process much more work at once. It directly attacks the bottleneck behind the scalability trilemma: that in a classic chain, every node must process everything. With sharding, each node only fully handles one shard, so the total capacity grows as you add shards, without asking any single machine to do more.

The term is borrowed from large databases, where a giant table is divided across many servers so no one server holds the whole thing. The hard part on a blockchain is keeping the shards secure and in agreement: if validators are spread thin across many shards, a single shard could be captured by an attacker who would never have the strength to take over the whole chain. Real designs counter this by randomly shuffling validators between shards and by giving every shard a shared anchor of security, so no slice can be quietly overpowered.

Imagine one overwhelmed clerk who must personally record every transaction in a country, against a system of regional offices that each handle their own area while reporting up to a national registry. The regional offices work in parallel and the country gets far more done, as long as the central registry keeps them honest and consistent. Sharding is a powerful idea, and modern designs increasingly pair a lighter form of it (sharding the data, not the execution) with rollups, so the base chain mainly guarantees that data is available for layer-2 systems to build on.