What is sharding?
- Author
- CRYPTO PORT Editorial
- Published
- Updated
- Reading time
- 5 min
In short
Sharding splits a network into partitions so that each node processes only its own slice rather than everything. Because it abandons the assumption that every node validates every transaction, cross-shard operations and security guarantees are hard, and implementations have moved cautiously.
Key points
- Partitions the network so work can run in parallel
- Each node stores only the data for its own partition
- Cross-shard consistency is the central difficulty
- Ethereum prioritised splitting data availability first
Definition
A design that divides a blockchain's data or execution into partitions called shards, with each node handling only some of them, so total capacity rises.
The idea comes from databases, where a table too large for one server is split across several by key range. Applied to a blockchain, it drops the rule that every node validates everything: each node validates only its shard, and capacity in principle grows with the number of shards.
The hard part is anything that crosses shards, such as moving an asset from one to another. Both states must update consistently; allow a half-completed transfer and the asset can be counted twice. Security is the other issue — if attacking one shard is cheaper than attacking the whole, that becomes the weak point.
Ethereum originally planned to shard execution itself, then reoriented around rollups and began by splitting data availability instead. The blob space introduced by EIP-4844 is a dedicated area that holds rollup data for a limited window, and it is the first step along that path.
Watch out for
- · More shards do not help workloads dominated by cross-shard activity
- · Fewer validators per shard makes attacking a single shard cheaper
- · Chains use the word for quite different designs, so do not compare on the label alone