Ethereum logo

GossipSub

Key metrics on the Ethereum pubsub network

Bandwidth consumption statistics, Message bandwidth consumption, Number of duplicates, Block arrival times.

Bandwidth

Bandwidth consumption breakdown per direction and message type

Inbound Outbound Bandwidth Usage

Inbound Outbound Bandwidth Usage

Shows the GossipSub bandwidth of our Hermes node in kilobytes per second (kB/s), averaged over 5-minute intervals, per direction:

  • In: messages and control messages that our node received, including duplicate copies of messages.
  • Out: messages and control messages that our node sent.
  • Drop: messages and control messages that our node dropped instead of sending them.

The size of the IHAVE, IWANT and IDONTWANT control messages is an estimate from the number of message IDs in them. Subscriptions, GRAFT and PRUNE messages are not included. On Ethereum, the data only includes the beacon_block, beacon_attestation_* and blob_sidecar_* topics. Vertical lines mark a new deployment of our Hermes node.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Bandwidth Usage in Kilobytes/s agg. in 5m intervals

Bandwidth Usage by Message Type

Bandwidth Usage by Message Type

Shows the GossipSub bandwidth of our Hermes node per message type in kilobytes per second (kB/s), averaged over 5-minute intervals:

  • Deliver: the first copy of a message that our node received and delivered after validation.
  • Duplicated: more copies of a message that our node already had.
  • Msg: messages that our node sent to its peers.
  • Ihave and Iwant: control messages. Their size is an estimate from the number of message IDs in them.

The direction menu shows only received (Inbound) or sent (Outbound) traffic, or all traffic (Any Direction). On Ethereum, the data only includes the beacon_block, beacon_attestation_* and blob_sidecar_* topics. Vertical lines mark a new deployment of our Hermes node.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Insights into the share of each messages bandwidth consumption

Bandwidth Usage Distribution

Bandwidth Usage Distribution

Shows the share of each message type or topic in the GossipSub bandwidth of our Hermes node over the latest 24 hours of data. Hover over a slice to see its share in percent.

  • By Type shows the message types: deliver is the first copy of a message that our node received, duplicated are more copies of a message that our node already had, msg are messages that our node sent, and ihave and iwant are control messages.
  • By Topic shows the GossipSub topics.

The second menu selects the direction: Inbound (received), Outbound (sent), or Any Direction, where each slice name starts with in or out. Control messages that our node dropped are not included. On Ethereum, the data only includes the beacon_block, beacon_attestation_* and blob_sidecar_* topics.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Loading...

Block Arrival Times

Block arrival times per geographic region (continent) for every client implementation and for different percentiles.

Block Arrival Times Filtered

Block Arrival Times Filtered

Shows the Min, Avg, P50 (median) and P95 of the Ethereum block arrival times per epoch (6.4 minutes). The block arrival time is the time in seconds between the start of the slot (when the block should be created) and the moment a sentry node of Xatu, the data collection of ethPandaOps, saw the block. Arrivals later than 12 seconds are not counted. The continent menu shows only the arrivals at sentry nodes on one continent.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Aggregated Block Arrival Times Filtered by Continent and Agent Type.

Block Arrival Times Grouped

Block Arrival Times Grouped

Shows one statistic of the Ethereum block arrival times, with one line per continent or per agent type. The block arrival time is the time in seconds between the start of the slot (when the block should be created) and the moment a sentry node of Xatu, the data collection of ethPandaOps, saw the block. Arrivals later than 12 seconds are not counted. Each value covers one epoch (6.4 minutes) for time ranges of up to 100 days, and one day for longer time ranges.

  • By Continent shows one line per continent of the sentry node.
  • By Agent Type shows one line per consensus client (agent type) of the sentry node. Grandine and Caplin are hidden at first. Click them in the legend to show them.

The second menu selects the statistic: Avg, Min, or a percentile from P25 to P95.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Aggregated Block Arrival Times Grouped by Continent and Agent Type.

Msg Arrival Times CDF

Msg Arrival Times CDF

Shows the distribution of Ethereum block arrival times on the selected day. The arrival time is the time in seconds between the start of the slot and the moment a sentry node of Xatu, the data collection of ethPandaOps, saw the block. Arrivals later than 12 seconds are not counted. The y-axis shows the share of arrivals that took this long or less.

  • By Continent shows one line per continent of the sentry node. The second menu shows only the sentry nodes that run one consensus client (agent type), or Any Agent.
  • By Agent Type shows one line per consensus client of the sentry node. The second menu shows only the sentry nodes on one continent, or Any Continent.

The tooltip shows the number of samples, which is the number of arrivals in each line. Use the date picker to select another day.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Loading ...

Message Size / Duplicate Count Correlation

Message Size / Duplicate Count Correlation

Shows how the arrival time of Ethereum blocks relates to their size on the selected day. Each cell shows the share of all block arrivals of the day, in percent, with this arrival time and block size. The x-axis groups the time between the start of the slot and the arrival of the block in steps of 0.5 seconds. The y-axis groups the compressed block size in steps of 50 kB. One arrival is one block seen by one sentry node of Xatu, the data collection of ethPandaOps. Arrivals later than 12 seconds are not counted. Use the date picker to select another day.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Relation between message size and the likelihood of receiving duplicates

Duplicates

Duplicate message statistics per topic, geographic distribution and client implementation

Duplicates Per Topic CDF

Duplicates Per Topic CDF

Shows the distribution of the number of duplicates per message for each topic on the selected day. A duplicate is one more copy of a message that our Hermes node already received. The x-axis shows how many duplicates of a message our node received. The y-axis shows the share of unique messages, identified by their message IDs, with this number of duplicates or fewer. Click a topic in the legend to show or hide it, and use the date picker to select another day.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Loading ...

Time to First Duplicate

Time to First Duplicate

Shows the distribution of the time to the first duplicate per topic on the selected day. This is the time in milliseconds from the moment our Hermes node delivered a message to the moment the first duplicate of this message arrived. A duplicate is one more copy of a message that our node already received. Only messages with at least one duplicate count. Messages whose first duplicate arrived before our node delivered the message, for example during validation, are not included. The x-axis shows the first 2 seconds. The tooltip shows the number of samples, which is the number of messages of each topic. Click a topic in the legend to show or hide it, and use the date picker to select another day.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

CDF of how long it took for the first duplicate to arrive.

Duplicate Message Sizes Per Topic CDF

Duplicate Message Sizes Per Topic CDF

Shows the distribution of message sizes per topic on the selected day. Each message that our Hermes node received and delivered counts once, by its message ID, also when it had no duplicates. The x-axis shows the size in kB (1 kB = 1,024 bytes). The y-axis shows the share of messages of this size or smaller. The tooltip shows the number of samples, which is the number of messages of each topic. Click a topic in the legend to show or hide it, and use the date picker to select another day.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Loading ...

Msg Size / Count Correlation

Msg Size / Count Correlation

Under the same network bandwidth, larger messages take longer to be fully transmitted between peers. Since a peer can only detect a message duplicate after downloading it completely, the likelihood of receiving duplicates increases with message size.

This heatmap shows how the size of the messages in one topic relates to the number of duplicates that our Hermes node received of them on the selected day. Each cell shows the number of unique messages with this size and this number of duplicates. The x-axis groups the message size in steps of 20 kB (1 kB = 1,024 bytes). The y-axis shows the number of duplicates: the row 2-3 holds the messages with 2 duplicates. The menu selects the topic, and the date picker selects the day.

Data is provided as is, without warranty. Scraping and automated bulk extraction are not permitted. See our Terms of Use.

Loading ...

Related Work

Blog posts, papers and talks by the ProbeLab team about Ethereum.

YouTube thumbnail: 7EnDYwcMGq4
Talk

devp2p, libp2p, ethp2p - Untangling Ethereum’s Networking Stack

Ethereum's execution layer runs devp2p: RLPx transport, capability negotiation, and request-response subprotocols carrying the mempool and sync. Its consensus layer runs libp2p, a modular general-purpose …

YouTube thumbnail: tB2k2axL6DE
Talk

What can Ethereum learn from other networks?

Ethereum's peer-to-peer layer was designed for a protocol that looked very different from the one we run today, and it did its job well. But …

ArXiv Preprint
Paper

Quantifying Ethereum Energy Consumption via Network Mapping

Ethereum’s electricity use fell by about 99.95% after the network moved from proof of work to proof of stake. Service providers still need to report …

From rawpixel.com / Freepik
Blog post

ProbeLab's 2025 Review and 2026 Outlook

What we've achieved in 2025 and what we're targeting in 2026

MIDDLEWARE '25 Conference Paper
Paper

PANDAS: Peer-to-peer, Adaptive Networking Allowing Data Availability Sampling within Ethereum Consensus Timebounds

Layer-2 protocols such as rollups can help address Ethereum's throughput limits. An efficient data availability layer is key for layer-2 support in Ethereum, but broadcast …

SIGMETRICS '26 Conference Paper
Paper

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

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 (P2P) …