Data partitioning (sharding) is a method of scaling a distributed ledger in which the blockchain is split into several parallel-working segments called shards. In the classic architecture each network node stores a full copy of the entire chain and verifies every transaction, which limits performance as the number of participants grows. Sharding removes this limitation by distributing data and computations among groups of nodes.

How sharding works

Each shard processes its own set of transactions and maintains part of the overall network state. Nodes no longer need to keep the entire ledger — it is enough to be responsible for their own segment. Through the parallel processing of many shards the network is able to perform significantly more operations per unit of time.

• increased throughput and processing speed;
• reduced storage and computation requirements for an individual node;
• the possibility of horizontal network growth as the load increases.

The key engineering task of sharding is to ensure the security and consistency of data between shards, as well as to correctly carry out operations that affect several segments at once. Therefore such solutions require well-thought-out mechanisms for shard interaction and protection against the concentration of dishonest nodes in a single segment.