Transaction latency testing tool

A Go-based tool for testing and comparing transaction latency performance between different Flashblocks-enabled Base node endpoints.

Flashblocks are sub-blocks issued by the block builder every 200ms, allowing for early confirmation times and making Base 10x faster for real-time applications.

Requirements

Node compatibility

For synchronous transaction sending testing, the node endpoints must support the eth_sendRawTransactionSync RPC method. See EIP-7966: eth_sendRawTransactionSync Method.

For asynchronous transaction sending testing, the node endpoints do not have to support the eth_sendRawTransactionSync RPC method and will rely on separate polling for confirmed transactions.

Chainstack Base nodes for Mainnet & Testnet support both the synchronous one with eth_sendRawTransactionSync and the asynchronous one.

Configuration

Create a .env file based on .env.example:

PRIVATE_KEY=your_private_key_here
TO_ADDRESS=0xc2F695613de0885dA3bdd18E8c317B9fAf7d4eba

# If you are going with the asynchronous method and poll for the transaction receipts separately
POLLING_INTERVAL_MS=50

BASE_NODE_ENDPOINT_1=https://your-node-endpoint-1.com
BASE_NODE_ENDPOINT_2=https://your-node-endpoint-2.com

# This a simple label for the resulting filename to remember where you sent the test transactions from. Not used in any node routing.
REGION=singapore

NUMBER_OF_TRANSACTIONS=10

# With `true`, the synchronous eth_sendRawTransactionSync method is used. With `false`, the asynchronous method is used with POLLING_INTERVAL_MS=50
SEND_TXN_SYNC=true

RUN_ENDPOINT2_TESTING=true

Configuration options

Transaction sending testing modes

SEND_TXN_SYNC=true:

• Uses eth_sendRawTransactionSync method
• Provides instant confirmation receipt when transaction is included in a Flashblock
• Transactions wait for immediate inclusion confirmation

SEND_TXN_SYNC=false:

• Uses standard eth_sendTransaction method
• Polls for transaction receipts using POLLING_INTERVAL_MS
• Traditional async transaction sending

Polling configuration

POLLING_INTERVAL_MS:

• Used only when SEND_TXN_SYNC=false
• Defines how frequently (in milliseconds) to check for transaction receipts
• Default: 50ms (recommended for accurate latency measurements)
• Lower values = more frequent polling = faster detection but higher load
• Not used when SEND_TXN_SYNC=true since confirmations are immediate

Endpoint testing

• BASE_NODE_ENDPOINT_1: Primary endpoint (e.g., a Flashblocks-enabled)
• BASE_NODE_ENDPOINT_2: Secondary endpoint (e.g., a standard non-Flashblocks endpoint)

The tool will:

1. Send transactions to BASE_NODE_ENDPOINT_1 first
2. Then send transactions to BASE_NODE_ENDPOINT_2 (if RUN_ENDPOINT2_TESTING=true)
3. Compare performance between both endpoints

Usage

# Build the container
docker build -t transaction-latency .

# Run the test
docker run -v $(pwd)/data:/data --env-file .env --rm -it transaction-latency

Note about .env file loading

When running with Docker, you may see the log message:

Error loading .env file

This is expected and harmless. The application uses two methods to load environment variables:

1. Docker's --env-file flag (recommended): Sets environment variables at container runtime
2. Direct .env file loading: Attempts to read the .env file from inside the container

Since we use the --env-file approach for security (keeping secrets out of the image), the direct file loading fails but the environment variables are still available via Docker. The application continues normally and all configuration works as expected.

Results

After completion, results are saved to the ./data/ directory:

• endpoint1-{region}.csv: Results from BASE_NODE_ENDPOINT_1
• endpoint2-{region}.csv: Results from BASE_NODE_ENDPOINT_2 (if enabled)

CSV format

Each CSV contains:

• sent_at: Timestamp when transaction was sent
• txn_hash: Transaction hash
• included_in_block: Block number where transaction was included
• inclusion_delay_ms: Time from sending to confirmation (milliseconds)

Performance expectations

Flashblocks are produced at 200ms intervals, but actual confirmation times will be higher due to network latency:

• Standard endpoint: ~2000ms (2-second block time)
• Flashblocks endpoint: ~300-500ms (200ms Flashblock interval + network travel time)

The actual confirmation time depends on:

• Network latency to/from the Base node
• Transaction processing time
• Time until next Flashblock (up to 200ms)
• Network travel time for confirmation response

Optional eth_getBlockByNumber comparison

On a Flashblocks-enabled node, the following standard RPC methods will retrieve the Flashblocks data instead of full blocks data:

• eth_getBlockByNumber with pending tag
• eth_getTransactionReceipt
• eth_getBalance with pending tag
• eth_getTransactionCount with pending tag
• eth_getTransactionByHash with pending tag
• eth_sendRawTransactionSync

For a quick comparison, see:

• eth-get-block-by-number-pending-examples/preconfirmed-flashblock.log — the result of running "method":"eth_getBlockByNumber","params":["pending",true] on a Flashblocks-enabled node as block 33228756 was forming.
• eth-get-block-by-number-pending-examples/confirmed-block.log — the result of running of running "method":"eth_getBlockByNumber","params":["pending",true] on a non-Flashblocks-enabled node as block 33228756 was formed.

You will see the key differences in the results:

• Transaction count — 52 in the Flashblock vs. 167 in the fully formed block.
• stateRoot — empty in the Flashblock vs. computed in the fully formed block.
• blockHash — different in the Flashblock and the fully formed block. The finalized transactions will have the hash of the fully formed block attributed to them.
• receiptsRoot — different in the Flashblock and the fully formed block.

Learn more

• Base Flashblocks documentation
• We’re making Base 10x faster with Flashblocks