Also covering BIP152

Compact block relay is a protocol that allows two nodes with roughly similar sets of unconfirmed transactions to minimize both the bandwidth and the latency required to transfer a block that confirms many of those same transactions.

By default, most full nodes are configured to receive relayed unconfirmed transactions, which the nodes store in their mempools and further relay to their other peers. Ultimately those transactions reach miners, who attempt to include them in a block. When that new block is itself relayed back to nodes, it typically consists almost entirely of transactions that the nodes already have in their mempools.

Relaying those transactions a second time is redundant, and it’s that redundancy that BIP152 compact blocks aims to eliminate. A sending node constructs a compact block by replacing each transaction it believes its peer has with a fast per-peer 6-byte non-cryptographic hash of that transaction’s identifier (txid for pre-segwit nodes, witness txid (wtxid) for segwit nodes). This can take a multimegabyte block with thousands of transactions and compact it down to just a few kilobytes. This directly reduces bandwidth. Indirectly, it also significantly reduces latency because TCP propagates smaller amounts of data faster than larger amounts.

For any transactions the generating node doesn’t think its peer knows about, it sends the full transaction.

The receiving peer hashes the wtxids of all the transactions in its mempool. It then substitutes each wtxid hash in the compact block with the full transaction from its mempool. If any of the wtxid hashes in the compact block don’t match a transaction in the receiving peer’s mempool, it requests that transaction from the sending node.

To maximize both bandwidth and latency improvements, compact blocks can be used in two different modes:

  • ● Low bandwidth mode is designed to be used with most peers. When a node learns about a new block, it announces the header hash of that block to its BIP152 low-bandwidth peers. Only if a peer requests the compact block is any additional data sent. This avoids sending data to peers who may have already received the block from one of their other peers.

  • ● High bandwidth mode is designed to only be used with a few peers who specifically request the mode. When a node receives a block, it verifies its header contains an appropriate amount of Proof of Work (PoW) and then, without performing any other significant verification, immediately sends it as a compact block to its high bandwidth mode peers. This allows blocks to propagate across the network quickly at the cost of nodes sometimes sending a compact block to a peer who already received the block from another node. Since compact blocks are so much smaller than raw blocks, this “high bandwidth” mode still typically uses less bandwidth than pre-BIP152 block relay.

BIP152 also specifies two versions of compact blocks:

  • Version 1 uses hashes of txids.

  • Version 2 uses hashes of wtxids.

Preserving the advantages of compact block relay is often cited as one reason for trying to keep mempool and relay policy consistent between nodes—the more the mempools of nodes differ, the less effective compact blocks will be at minimizing bandwidth and latency.

Primary code and documentation

Optech newsletter and website mentions

2022

2021

2020

Previous Topic:
Compact block filters Next Topic:
Consensus cleanup soft fork

Edit page
Report Issue