Bitcoin Cluster Mempool

Overview

Bitcoin Cluster Mempool is an architecture used by Bitcoin Core to organize related unconfirmed transactions inside the mempool. A cluster is formed from transactions connected by unconfirmed spending relationships. If transaction B spends an output created by transaction A, they belong to the same dependency graph. Grouping these relationships explicitly allows the node to reason about the combined fee incentives of parents, children, and larger packages rather than evaluating every transaction as an isolated item.

The design divides clusters into fee-ranked chunks. A chunk can contain one or more transactions that should be considered together because a low-fee parent may only be economically attractive when paired with a higher-fee child. The resulting feerate diagram supports more consistent choices for mining, mempool eviction, relay, and replacement. It also reduces edge cases created by older ancestor and descendant accounting rules, where different algorithms could rank the same package differently.

Cluster mempool is a node implementation and policy mechanism. It does not alter the consensus validity of a blockchain transaction or require every node to hold an identical set of pending transactions. A node can reject, replace, or evict an unconfirmed transaction from its local pool while still accepting a later block that validly includes it. Wallets and payment systems should therefore distinguish mempool visibility from network-wide acceptance and from block confirmation.

For developers, the practical effects appear in fee estimation, package submission, Replace-by-Fee behavior, and diagnostic RPCs that describe clusters or fee diagrams. Applications that depend on specific ancestor limits or replacement behavior should test against current Bitcoin Core policy instead of assuming historical mempool rules. The architecture is particularly relevant to protocols that use pre-signed transactions, child-pays-for-parent fee bumping, or tightly connected transaction packages.