Peer Selection

Nodes prioritize connections to low-latency peers to minimize propagation delay. This is often measured via ping time or round-trip time (RTT). Geographic proximity is a common heuristic, but network topology (e.g., internet backbone routes) is the true determinant. Faster peers ensure quicker block and transaction dissemination, reducing the risk of forks.

A robust client connects to a diverse set of peers to avoid sybil attacks and eclipse attacks. Key diversity factors include:

• Client Implementation: Connecting to peers running Geth, Erigon, Nethermind, etc.
• Network Topology: Ensuring peers are on different autonomous systems (ASNs).
• Geographic Distribution: Avoiding concentration in a single region. This prevents a single point of failure or manipulation.

Nodes maintain a reputation score for each peer based on historical behavior. Positive actions (timely delivery of valid blocks) increase the score, while negative actions (propagating invalid transactions, frequent disconnects) decrease it. Peers with low scores are disconnected or deprioritized. This system, often called peer scoring, is formalized in protocols like libp2p's gossipsub.

Peers must speak compatible wire protocols (e.g., Ethereum's devp2p, libp2p) and support the same network ID (Mainnet, Goerli). Version mismatches cause incompatibility. Nodes often implement handshake protocols to negotiate supported capabilities (like snap sync) upon connection, only maintaining links with peers that support required features.

Nodes enforce strict limits on connection counts for stability.

• Outbound Peers: Actively initiated connections. A client (e.g., Geth) defaults to ~50 outbound peers to actively pull data from.
• Inbound Peers: Passive connections accepted from others. Limits (e.g., ~100) prevent resource exhaustion. These limits are configurable and balance resource usage with network participation.

The initial peer set is found via bootnodes—hardcoded entry points—and peer discovery protocols. Key mechanisms include:

• DNS-based Discovery: Querying DNS TXT records for peer lists.
• Discv5: Ethereum's Node Discovery Protocol v5, enabling encrypted, topic-based peer finding.
• Peer Exchange: Requesting peer lists from already-connected nodes.

Nodes prioritize peers with low network latency and high historical uptime to ensure fast, reliable data propagation. This is critical for minimizing block propagation delays and preventing network partitions. Metrics like ping time and connection duration are continuously evaluated.

Peers must run compatible protocol versions to communicate effectively. Nodes check version strings and network magic numbers. A key function is identifying and avoiding peers on a different chain fork, which prevents the propagation of invalid or stale blocks and transactions.

To prevent centralization and increase resilience, clients often seek geographically distributed peers and peers from different autonomous systems (ASNs). This reduces the risk of correlated failures (e.g., a regional ISP outage) and improves the censorship-resistance of the peer-to-peer graph.

Nodes manage resource consumption by setting limits on inbound and outbound connections. High-bandwidth peers capable of serving blocks quickly are valued. Clients enforce maximum peer counts and may implement sybil resistance mechanisms, like requiring proof-of-work for incoming connections.

Many clients maintain a dynamic reputation score for each peer based on behavior. Positive actions (relaying valid blocks) increase the score, while negative actions (sending invalid data, spamming, dropping connections) lead to penalties or banning. This creates a self-policing network.

Peers advertise capabilities via service flags. The most important is NODE_NETWORK, indicating a full node that can serve the entire blockchain history. Nodes may prefer these peers over lightweight clients (SPV nodes) that only request filtered data, ensuring access to full validation data.

A security property that peer selection algorithms must consider. It ensures the network cannot be easily dominated by an attacker creating many fake identities (Sybils). Selection strategies that incorporate Proof-of-Work, stake, or reputation scores help mitigate Sybil attacks by making identity creation costly or by penalizing malicious behavior.

A network-level attack where a malicious actor isolates a target node by monopolizing all its peer connections. The attacker uses the peer selection process to become the target's only source of blockchain data, enabling double-spends or censorship. Robust peer selection with randomized and diverse peer sets is a primary defense.