Skip to content
Crypto.world

Research

Multiple Sides of 36 Coins: Measuring Peer-to-Peer Infrastructure Across Cryptocurrencies

The networking layer that carries transactions and blocks between blockchain nodes is largely unmeasured for all but Bitcoin and Ethereum, even though network size and concentration are increasingly…

Multiple Sides of 36 Coins: Measuring Peer-to-Peer Infrastructure Across Cryptocurrencies
arXiv
2511.15388
Published
2025-11-19
Authors
Lucianna Kiffer, Lioba Heimbach, Dennis Trautwein, Yann Vonlanthen, Oliver Gasser

AI summary

Overview

Research area: Blockchain peer-to-peer (P2P) networking and Internet-scale network measurement.

Technical level: Intermediate. The paper assumes familiarity with blockchain fundamentals and with peer discovery concepts such as address gossip and Kademlia DHTs, but its findings are presented as network statistics that a general technical reader can follow.

Scope: A nine-month, multi-chain measurement study of the P2P layers of 36 public blockchain networks, using active crawlers, connectivity probes, Internet-wide scans and node-explorer scraping to compare network size, reachability, centralization and churn.

What This Paper Is About

The networking layer that carries transactions and blocks between blockchain nodes is largely unmeasured for all but Bitcoin and Ethereum, even though network size and concentration are increasingly used as evidence of resilience or risk. The authors build and operate a measurement pipeline across many chains at once to answer a basic question that no one has systematically answered: how large, how reachable and how decentralized are these P2P networks in practice? They combine crawlers, liveness probes, third-party explorer data and Internet-wide port scans into a single comparative framework.

Key Contributions

  1. First longitudinal, cross-network measurement of 36 public blockchains. The authors deployed 15 active crawlers, sourced data from community crawlers (two according to the abstract, one according to the introduction), and ran hourly connectivity probes over a study window described in the abstract as more than 9 months since late 2024 and in the results discussion as an eight-month window.

  2. Inference of 19 additional networks from shared discovery layers. By analyzing Ethereum's discv4 and discv5 discovery data and Bitcoin Cash discovery data, the authors inferred the existence and metadata of 19 auxiliary networks that reuse those discovery protocols rather than running their own.

  3. A generalizable Internet-wide scanning method. Instead of implementing custom discovery and handshake logic per chain, the authors probe each protocol's default ports with a network-specific payload using ZMap and ZGrab2. They validate the approach on Bitcoin and similar networks against known ground truth, then apply it to Cardano, which they could not crawl directly.

  4. A reusable measurement framework and public artifacts. The paper reports that measurement code will be made publicly available, points to a live dashboard at p2pobservatory.networks.imdea.org, releases the DHT crawler as Nebula on GitHub, and offers data access by request for academic use.

Main Findings

  • Network size spans orders of magnitude. Active node counts range from under 10 to more than 10,000 across the studied networks (the introduction describes the range as tens to tens of thousands). Ethereum discv4 and discv5 consistently register over 20,000 active nodes, with discv4 exceeding 30,000 at its peak; Bitcoin and Polkadot remain in the same order of magnitude; Avalanche, Solana and Avail sustain a few thousand peers; Litecoin, Stellar, Near, Ripple, Dogecoin, Bitcoin Cash (*), Filecoin and Celestia operate at only a few hundred.

  • Reachability varies sharply. Some networks exhibit fewer than 10% reachable nodes. The median discovery-hit ratio (active divided by discovered) is near unity for Polkadot (0.97), Filecoin (0.99), Celestia (0.96) and Avail (0.99), but below 0.05 for Bitcoin (0.04), Dogecoin (0.05), Bitcoin Cash (*) (0.03) and Litecoin (0.03). Ethereum discv4 and discv5 fall in between at 0.57 and 0.27.

  • Discovered address space is often enormous relative to live nodes. Median discovered IPs are 257,019 for Bitcoin and 28,063 for Litecoin against median active counts of 11,309 and 952 respectively; Ethereum discv4 reports 57,806 discovered and 33,318 active.

  • Hosting concentration is high on several chains. Ripple, Solana, Stellar and Near have hosting-provider proportions above 0.90 (Solana 0.96, Near 0.98, Stellar 0.94, Ripple 0.90). Bitcoin has the highest ISP share at 0.46.

  • IPv6 adoption is limited. Bitcoin-based networks lead with roughly 15–20% IPv6 usage (Bitcoin 0.15, Dogecoin 0.18, Bitcoin Cash (*) 0.19, Litecoin 0.20); notable exceptions elsewhere are Polkadot (0.20) and Celestia (0.16). Ethereum discv4 reports 0.00 IPv6 and Solana 0.00.

  • Probe type matters for liveness. TCP pings achieve median success above 0.80 for most networks, with exceptions Polkadot (0.19), Ethereum discv4 (0.66), Ethereum discv5 (0.62) and Avail (0.18). Bitcoin-client crawlers using protocol pings see median response above 0.80 (Bitcoin 0.87, Dogecoin 0.92, Bitcoin Cash (*) 0.82, Litecoin 0.90). Nebula DHT crawl pings succeed above 0.90 for Ethereum discv4 (0.91), Ethereum discv5 (0.90), Polkadot (0.93), Filecoin (0.97), Celestia (1.00) and Avail (0.99). Bitcoin-based networks respond more to TCP pings; DHT networks respond more to crawl pings.

  • Discovery coverage can be extremely concentrated. In some networks fewer than five nodes need to be contacted to discover 90% of reachable peers, while other networks exhibit much flatter coverage.

  • Shared discovery layers blur network boundaries. Bitcoin Cash and Bitcoin SV share the same network magic string, so they appear as one network at the discovery layer until the version handshake; the authors also identify a third protocol, eCash, in the same data. Ethereum discv4 encapsulates many other networks, which explains its much larger counts relative to discv5.

  • Node counts are stable over time. Despite the large differences between chains, node counts remain remarkably stable over the measurement window. A temporary dip in the Bitcoin-based networks in March came from a week-long IPv6 block due to a misconfiguration.

  • Coverage of the market is broad. The authors cover 8 of the top 10 protocols by market capitalization at the time (Binance is number 5 and is covered within the discv4 breakdown), and 14 of the top 21 (all above $5B).

Methodology in Plain English

The team ran three complementary measurement approaches from AWS infrastructure in Virginia (us-east-1).

Crawling. A crawler starts from known bootstrap nodes, asks each node for the peers it knows about, and repeats until no new addresses appear. For Bitcoin-family chains they use a crawler built on the Bitcoin Core client and the getAddress message, supplying the correct network magic (Bitcoin 0xd9b4bef9, Dogecoin 0xc0c0c0c0, Bitcoin Cash 0xe8f3e1e3, Litecoin 0xdbb6c0fb), with up to 10 retry attempts and a 20-second timeout. For Kademlia-based chains they use a generic crawler (Nebula) that crafts FIND_NODE queries to enumerate whole routing tables bucket by bucket across discv4, discv5 and libp2p. For Ripple, Avalanche and Near they exploit built-in RPC calls that return a node's current active connections, which is simpler but reveals nothing about the discovery process.

Liveness checks. Because a node can be online without answering a crawl query, they treat a node as active if it responds to any of several hourly probes: an in-protocol PING or status message, a TCP ping to the learned port and the protocol's default ports, or a completed crawl handshake. Default ports used include 8333 (Bitcoin and Bitcoin Cash), 22556 (Dogecoin), 9333 (Litecoin), 2459 and 51235 (Ripple), 9650 and 9651 (Avalanche), 3030 and 24567 (Near), 11625 (Stellar), and 8000 and 8001 (Solana). DHT networks were fully crawled every two hours.

Explorer scraping. They scraped six node explorers hourly, including Blockchair (IPv4 only), Bitnodes (IPv4, IPv6 and Onion), Ethernodes, ETCnodes, OBSRVR Radar and SolanaBeach, to cross-check crawler coverage.

Internet-wide scanning. For networks without usable discovery, they scan the full IPv4 space with ZMap and ZGrab2 using protocol-specific payloads, use the IPv6 Hitlist for likely-responsive IPv6 addresses, and limit the rate to 50kpps while following measurement best practices (blocklist, public description of the experiments).

Limitations the authors state. Running everything from one cloud region risks IP-based filtering and rate limits aimed at cloud ranges, and latency can cause timeouts; measuring actively means nodes behind NATs or firewalls are missed, so active counts are a lower bound. They note prior estimates that unreachable Bitcoin nodes are twice the reachable count. Solana's explorer API stopped returning IP-level information in mid May 2025, and Avail and Polkadot had a longer crawler disruption at the end of the study.

Why This Matters

Impact on research. Prior blockchain network measurement has concentrated almost entirely on Bitcoin and Ethereum. This paper provides a comparable baseline across dozens of chains, plus a reusable methodology (crawlers, RPC-based enumeration, port scans) that others can extend or replicate. It also shows why single-chain results should not be generalized: the same metric, such as reachability or IPv6 share, varies by an order of magnitude between ecosystems.

Real-world applications:

  • Regulatory and risk assessment. Frameworks such as EU MiCA activity thresholds and U.S. decentralization proposals rely on notions of scale and control; this work supplies concrete, comparable measurements of network size and hosting concentration.
  • Protocol design and client engineering. Knowing where nodes sit, how much IPv6 they use, and how concentrated discovery is informs decisions about peer table sizes, bootstrapping and address-relay design.
  • Resilience and attack-surface analysis. Discovery-hit ratios and AS-level concentration indicate how exposed a chain is to partitioning or eclipse-style attacks and how much redundancy its discovery layer actually provides.
  • Monitoring and benchmarking. The dashboard and annual-style re-measurement enable tracking whether decentralization claims hold over time.

Industry relevance. Node explorers, infrastructure providers, exchanges and staking operators can use these measurements to sanity-check their own views of network health. The finding that a handful of queried nodes can reveal 90% of reachable peers in some networks is directly relevant to anyone running monitoring or indexing infrastructure.

Future Directions

  • Continuous monitoring and benchmarking. The authors explicitly frame their methodology as opening the door to continued monitoring and more transparent assessments of blockchain infrastructure; the live dashboard is a first step.
  • Extending beyond reachable nodes. Because active measurement only captures nodes accepting inbound connections, better methods for estimating NATed and firewalled populations, building on the earlier Bitcoin estimate that unreachable nodes are roughly twice the reachable count, remain open.
  • Attribution within shared discovery layers. Decomposing multiplexed discovery networks such as discv4 into their constituent chains is only partially solved here; the truncated content does not report a complete attribution method for all shared layers.
  • Multi-vantage-point measurement. The authors identify their single AWS region as a bias source; repeating the study from diverse networks and geographies would test whether apparent centralization and reachability patterns hold.

Target Audience

Researchers and graduate students in blockchain systems, distributed systems and Internet measurement; protocol designers and client developers who need empirical grounding on peer discovery and network size; regulators and risk analysts assessing decentralization and concentration claims; and infrastructure operators, node explorers and monitoring teams who want a comparative picture of P2P network health across ecosystems. Readers looking for deep results on churn, geographic distribution or AS-level concentration should note that those analyses are referenced in Section 6 but their specific numbers are not present in the provided excerpt.

Authors’ abstract

Blockchain technologies underpin an expanding ecosystem of decentralized applications, financial systems, and infrastructure. However, the fundamental networking layer that sustains these systems, the peer-to-peer layer, of all but the top few ecosystems remains largely opaque. In this paper, we present the first longitudinal, cross-network measurement study of 36 public blockchain networks. Over 9 months, we deployed 15 active crawlers, sourced data from two additional…

The full abstract on arXiv
Read the original paper