geth_peri: A Go repository from WeizhaoT

Customized Go Ethereum for Latency Measurement

Customized Go-Ethereum client supporting Peri and latency measurement on Ethereum mainnet and testnets.

The faryar branch is customized based on the last commit on Jan 7, 2022 of go-ethereum.
The nemata branch is customized based on v1.10.25-stable release of go-ethereum.
- This branch is under construction and requires testing post-merge

See the original README documentation for other details.

Setup

This project requires a Go (version 1.14 or later), a C compiler, Python 3.9 or later, and node.js.

Install ntp server. On Linux, install the following executables.

sudo apt-get update 
sudo apt-get install -y ntp ntpdate

Install the following python packages.

pip install ntplib schedule web3

Install the following node.js packages.

npm install web3 fs-extra net path axios colors abi-decoder events sprintf-js async-mutex

Once the dependencies are installed, run

make geth

Running `geth` on Ethereum P2P network

Running the Node

To run a node on the mainnet with maximum peer count 50, execute the command

./build/bin/geth --mainnet --maxpeers 50 --datadir path_to_chaindata \
    --cache 12288 --ws --ws.port 8546  --ipcpath path_to_ipc \
    --loggypath path_to_loggy/loggy_example.json \
    --peri path_to_peri/peri_example.json

If the host machine does not have 12GB RAM, try to reduce the --cache argument in MB. Variants of this command are packed into shell scripts in

SHELL_SCRIPTS/
    |- mlistener.sh   #  listener nodes on mainnet for global latency measurement
    |- mtlistener.sh  #  listener nodes on mainnet for targeted latency measurement
    |- rtlistener.sh  #  listener nodes on Ropsten for targeted latency measurement
    |- rttransfer.sh  #  victim nodes on Ropsten for targeted latency measurement
    |- ktlistener.sh  #  listener nodes on Rinkeby for targeted latency measurement
    |- kttransfer.sh  #  victim nodes on Rinkeby for targeted latency measurement

Access to `geth` Console

When the node starts running, run the following command to get access to the javascript console of geth. This is used for tracking peers on-the-fly or checking synchronization progress.

./build/bin/geth attach path_to_ipc

This is packed into shell script SHELL_SCRIPTS/check.sh.

Loggy Configuration

The config of logger is fed to the --loggypath argument as the path to a json config file. Here is an example of path_to_loggy/loggy_example.json (See example file).

{
    "FlagLoggy": false, 
    "FlagConn": true,
    "FlagConnWarn": false,
    "FlagPerigee": true,
    "FlagForward": true,
    "FlagAllTx": false,
    "FlagObserve": true,
    "FlagBroadcast": false,
    "EPOCH_DURATION": 3600,
    "LOGS_BASEPATH": "/data/logs/mainnet"
}

The meaning of the fields are listed below:

Field	Description (if is a flag, when `true`)
`FlagLoggy`	[DEBUG] placeholder; set to `false`
`FlagConn`	record the peer connections
`FlagConnWarn`	warn about peer disconnections on the command line
`FlagPerigee`	record the Peri report on peer scores and actions
`FlagForward`	not block a relevant transaction from being forwarded
`FlagAllTx`	record late deliveries too (instead only the first deliveries of txs)
`FlagObserve`	record transactions; when `false`, no transaction is recorded
`FlagBroadcast`	record transaction broadcasting activities
`EPOCH_DURATION`	number of seconds before the logger creates a new file for logging
`LOGS_BASEPATH`	directory to the logs

Peri Configuration

The config of logger is fed to the --peri argument as the path to a json config file. Here is an example of path_to_peri/peri_example.json (See example file).

{
    "Active": false,
    "Targeted": false,
    "ShowTxDelivery": false,
    "Period": 1200,
    "ReplaceRatio": 0,
    "DialRatio": 3,
    "MaxDelayPenalty": 30000,
    "ObservedTxRatio": 4,
    "TargetAccountList": [],
    "NoPeerIPList": ["34.228.74.149", "35.172.193.247", "18.207.127.213", "3.82.23.217"]
}

The meaning of the fields are listed below:

Field	Description
`Active`	global switch of Peri
`Targeted`	global latency if `false`, targeted latency if `true` (target accounts in `TargetAccountList`)
`ObservedTxRatio`	(1 / sampling rate) of global latency; must be int
`ShowTxDelivery`	[DEBUG] print log of relevant transaction arrivals to command line if `true`
`Period`	Peri period length in seconds
`ReplaceRatio`	proportion of peers to be replaced
`DialRatio`	(1 / percentage of outbound peers). Must be int. 1 for Peri/Hybrid and 3 for Baseline
`MaxDelayPenalty`	$\bar \Delta$ in milliseconds
`TargetAccountList`	list of target accounts
`NoPeerIPList`	node with IP addresses in this list are permanently blocked

Running Script to Send Transactions

First, make sure the nodes are running and the accounts are unlocked. See official documentation for details.

Then, run this command to transfer 1 Gwei every 2 minutes from each unlocked account.

node BOT/scripts/transfer.js 2

Note: It may be necessary to change the gas price oracle in the scripts.

Running NTP Queries to Get Clock Offsets

Run the following command to keep recording clock differences between the host machine and multiple external IP addresses (for example, 1.1.1.1 and 2.2.2.2).

python SHELL_SCRIPTS/ntpm.py --logpath path_to_ntp_logs --ip 1.1.1.1 2.2.2.2

BloXroute

Check how to deploy a gateway and connect to a full node in the official website. It requires that --ws and --ws.port are set in geth command line arguments.

Experiment Data

The data consists of two parts: geth-logged and NTP clock differences. The former is under directory LOGS_BASEPATH, specified by the loggy configuration. The latter is under path_to_ntp_logs. These log files are sufficient for data analysis.

WeizhaoT/geth_peri