Nethermind Documentation

Pruning in Nethermind optimizes disk space by removing obsolete historical data while maintaining the current world state (accounts, contracts, etc.). This is the opposite of an archive node, which retains all history. Nethermind uses a 'hybrid pruning' approach by default, combining two methods:

1. In-memory pruning: A continuous process that keeps state changes in memory until a threshold is reached, reducing disk writes and improving block processing performance.
2. Full pruning: A thorough examination of the state tree that identifies and removes historical data, duplicating the current state to a new database and then replacing the old one. This significantly reduces database size back toward its initial state.

Starting with Nethermind 2.0, archive functionality is integrated into the flat database. Historical queries are answered using per-block state changesets captured during ordinary syncing. This removes the need for a separate archive database or a specialized sync mode. Archive nodes can be established either via a full sync from genesis or a snap sync to the tip that captures history from the pivot onwards.

Archive configuration is split into two independent groups:

1. FlatDb.History* settings: Govern historical state (used for eth_call, eth_getBalance, and eth_getStorageAt at historical blocks).
2. History.* settings: Govern historical blocks and receipts (used for eth_getBlockBy*, eth_getTransactionReceipt, and eth_getLogs).

Nethermind uses the JSON-RPC 2.0 protocol for interaction. It supports three primary transports, each suited for different use cases:

• HTTP: Idempotent, closes connections between requests; ideal for a low volume of requests with lower overall overhead.
• WebSocket: Continuous open channel; supports event subscriptions, streaming, and high volumes of requests with low per-message overhead.
• IPC (Unix domain socket): Most secure as it is limited to local interactions and cannot be exposed to external traffic; supports event subscriptions.