We’re excited to announce that a handful of the Xatu Consensus Layer P2P tables are now publicly available in Parquet format!
All data is captured directly from the Consensus Layer P2P network by utilising Probelab’s Hermes project and Xatu. Instances are run by the EthPandaOps team in 2 regions.
Summary:
- The data is dedicated to the public domain under the CC BY 4.0 license
- Data is partitioned by hour or day in Apache Parquet files
- We’ve been collecting this data for the last few months!
By providing this data we hope to enable more in-depth analysis and research into the Ethereum network’s behavior at the P2P level.
Tables #
About this data #
This data is captured closer to the wire in comparison to the those collected via the beacon api event stream. It is therefore not verified in any way. This means that messages with invalid signatures or other issues are not filtered out. This allows for a more comprehensive view of the network, including potential malicious or faulty behavior that might be filtered out at higher levels, but sacrifices some data quality.
Here’s a simplified diagram that shows how a block proposal would land in both the beacon api and the libp2p tables:
Getting started #
The easiest way to get started is to install the Clickhouse client and use clickhouse local
to query the data.
Working with the data has more details.
Examples #
The following examples demonstrate how these newly available tables can be utilized both independently and in conjunction with our existing Xatu tables to address intriguing questions about the Ethereum network.
Example 1 #
Which 3 slots had the fastest arriving block proposals on 15th of August on Mainnet?
Answer: We can use the libp2p_gossipsub_beacon_block
table to find the first seen time for each slot and then order by that value.
clickhouse local -q "
SELECT
min(propagation_slot_start_diff) as slot_time_ms,
slot
FROM
url('https://data.ethpandaops.io/xatu/mainnet/databases/default/libp2p_gossipsub_beacon_block/2024/8/15.parquet', 'Parquet')
GROUP BY
slot
ORDER BY
slot_time_ms
LIMIT 3
FORMAT PrettyCompact
"
┌─slot_time_ms─┬────slot─┐
│ 209 │ 9738557 │
│ 229 │ 9742293 │
│ 233 │ 9745006 │
└──────────────┴─────────┘
Example 2 #
Which blocks were seen before the deadline but still didn’t make it in to the finalized chain on Mainnet?
Answer: We can check the libp2p_gossipsub_beacon_block
table for blocks that were seen before the deadline on the p2p network, and then check the canonical_beacon_block
table for blocks that were finalized.
clickhouse local -q "
WITH p2p_blocks AS (
SELECT
block AS block_hash,
min(propagation_slot_start_diff) as first_seen_ms,
slot_start_date_time,
slot
FROM
url('https://data.ethpandaops.io/xatu/mainnet/databases/default/libp2p_gossipsub_beacon_block/2024/8/15.parquet', 'Parquet')
GROUP BY
block, slot_start_date_time, slot
HAVING
first_seen_ms < 4000
),
canonical_counts AS (
SELECT
block_root AS block_hash,
COUNT(*) AS count
FROM
url('https://data.ethpandaops.io/xatu/mainnet/databases/default/canonical_beacon_block/2024/8/{14..16}.parquet', 'Parquet')
GROUP BY
block_root
),
blocks_with_zero_count AS (
SELECT block_hash
FROM canonical_counts
WHERE count = 0
)
SELECT
p2p.slot as slot,
p2p.first_seen_ms
FROM p2p_blocks p2p
LEFT JOIN canonical_counts c ON p2p.block_hash = c.block_hash
LEFT JOIN blocks_with_zero_count b ON p2p.block_hash = b.block_hash
WHERE COALESCE(c.count, 0) = 0
ORDER BY p2p.first_seen_ms ASC
LIMIT 10
FORMAT PrettyCompact
"
┌────slot─┬─p2p.first_seen_ms─┐
│ 9743140 │ 924 │
│ 9743136 │ 3053 │
│ 9738789 │ 3476 │
│ 9738302 │ 3725 │
│ 9743407 │ 3837 │
└─────────┴───────────────────┘
Example 3 #
How long did it take for attestations from validator 0 to be included on chain in the Sepolia network, and when were they first seen on the p2p network?
Answer: We can compare the time attestations were seen on the Sepolia p2p network with the time they were included on chain. We’ll use the libp2p_gossipsub_beacon_attestation
table for p2p data and the canonical_beacon_elaborated_attestation
table for when they were included on chain, joining them based on the block root.
Note: Attestation files can be large so we’re only looking at an hour of attestation data.
clickhouse local -q "
WITH attestations AS (
SELECT
slot,
min(propagation_slot_start_diff) as first_seen_ms
FROM
url('https://data.ethpandaops.io/xatu/sepolia/databases/default/libp2p_gossipsub_beacon_attestation/2024/8/8/0.parquet', 'Parquet')
WHERE attesting_validator_index = 0
GROUP BY slot
),
on_chain_attestations AS (
SELECT
slot,
min(block_slot) as first_seen_slot
FROM
url('https://data.ethpandaops.io/xatu/sepolia/databases/default/canonical_beacon_elaborated_attestation/2024/8/8.parquet', 'Parquet')
WHERE has(validators, 0)
GROUP BY slot
)
SELECT
a.slot,
a.first_seen_ms as p2p_first_seen_ms,
o.first_seen_slot - a.slot AS inclusion_delay_slots
FROM attestations a
LEFT JOIN on_chain_attestations o ON a.slot = o.slot
ORDER BY a.slot ASC
FORMAT PrettyCompact
"
┌────slot─┬─p2p_first_seen_ms─┬─inclusion_delay_slots─┐
│ 5611821 │ 1036 │ 1 │
│ 5611852 │ 885 │ 1 │
│ 5611890 │ 4188 │ 1 │
│ 5611920 │ 701 │ 1 │
│ 5611960 │ 4145 │ 2 │
│ 5611985 │ 580 │ 1 │
│ 5612027 │ 4153 │ 2 │
│ 5612037 │ 764 │ 1 │
│ 5612067 │ 534 │ 1 │
└─────────┴───────────────────┴───────────────────────┘
Wrapping up #
We hope you find this data useful and look forward to seeing what you build with it! If you have any questions or feedback please reach out to us on Twitter or join the Xatu Telegram Group
Happy querying!
Love,
EthPandaOps Team ❤️