Data & code for Empirical Analysis of EIP-1559

This repository holds the data and code used for the paper "Empirical Analysis of EIP-1559: Transaction Fees, Waiting Time, and Consensus Security" (ACM CCS 2022; online version here).

Summary

• Waiting time (Section 5.3): Folder blockdata/ contains waiting time data in blockdata.npy and waitingtime_csv.csv.
• Network spikes (Table 4 & Figure 10): Folder spikedata/ contains the data of network spikes.
• Miner's revenue (Figure 11 & 12): Folder MEVdata/ contains data used for MEV analysis. Folder MEVfig/ contains a detailed MEVdata.csv file and the figures.

Reproducing data & graphs

Raw mempool data

Raw mempool data used to calculate waiting time can be downloaded here (about 20GB). After downloading it you should create a folder named compressed/ to store the uncompressed raw data.

You can see several .txt files, containing the timestamp of transactions received by a particular full node. For example, the file LA_[2021070100,2021071600)_compressed.txt contains timestamp collected by our full node in LA between July 1, 2021 and July 16, 2021. Each line of these files describes a transaction. E.g., the following line means that transaction 0x16... is received at Unix timestamp 1626418821.870.

0x16d1f71ef96c9456dc465ee2ce4d106b0dda0d440380c1cdd053c9deb58d8284 1626418821.870

blockdata/

The script waitingtime.py processes the raw data and generates two files: the block information database blockdetail.npy, and a detailed transaction waiting time table. Outputs are stored in the blockdata/ folder.

In the same folder there are three other files gas_csv.csv, timestamp_csv.csv and sibling_csv.csv, recording the gas usage, timestamp, and sibling count of the blocks respectively, derived from the Ethereum blockchain. These files are used to analyze network spikes in the paper.

spikedata/

The script spike.py reads gas_csv.csv, timestamp_csv.csv, and sibling_csv.csv in blockdata/ and outputs avggas.csv (in spikedata/ folder). The file avggas.csv tells the average gas per second around each block. From this file, we can obtain Figure 10 and Table 4.

MEVdata/

MEVdata contains the dataset for miner revenue, including Flashbots revenue collected from the Flashbots API, as well as other sources of rewards (including block rewards, uncle rewards, etc).

To reproduce the data, you can use the following command.

./mev.py --data

When executing the --data command, mev.py will call the Flashbots API, combining data from Ethereum full nodes (via web3_api.py) to generate MEV raw data. This data will be saved in the MEVdata/ folder.

MEVfig/

The script mev.py plots data in MEVdata/. The figures (figure 11 and figure 12 of our paper) are saved in the MEVfig/ folder. To reproduce the figure, you can use the following command.

./mev.py --csv
./mev.py --img

Specifically, the first command will read the MEV raw data in the MEVdata folder and output a table MEVfig/MEVdata.csv with all the data. The second command will to draw figures using MEVfig/MEVdata.csv.