Yellowstone Thorofare

Benchmarking tool for Solana Geyser gRPC endpoints. Compares two endpoints simultaneously to measure actual performance differences.

Build

cargo build --release

What it measures

• All 6 slot stages: FirstShredReceived, CreatedBank, Completed, Processed, Confirmed, Finalized
• Download time (FirstShredReceived > Completed)
• Replay time (CreatedBank > Processed)
• Account update propagation delays (optional)
• Transaction propagation delays (optional)

Basic usage

thorofare --endpoint1 endpoint1.com:10000 --endpoint2 endpoint2.com:10000

Arguments

Required:

• --endpoint1 - First endpoint (host:port)
• --endpoint2 - Second endpoint (host:port)

Optional:

• --x-token1 - Auth token for endpoint1
• --x-token2 - Auth token for endpoint2
• --slots - Number of slots to collect (default: 1000)
• --with-accounts - Track account updates
• --account-owner - Filter by program ID (requires --with-accounts)
• --with-transactions - Track transaction updates (non-vote)
• --endpoint1-richat - Use Richat interface for endpoint1
• --endpoint2-richat - Use Richat interface for endpoint2
• --config - Config file (default: config.toml)
• --output - Output file (default: benchmark_results.json)
• --log-level - Log level (default: info)

Examples

With auth tokens:

thorofare \
  --endpoint1 endpoint1.com:10000 \
  --endpoint2 endpoint2.com:10000 \
  --x-token1 YOUR_TOKEN_1 \
  --x-token2 YOUR_TOKEN_2

Track account updates by program owner:

thorofare \
  --endpoint1 endpoint1.com:10000 \
  --endpoint2 endpoint2.com:10000 \
  --with-accounts \
  --account-owner 675kPX9MHTjS2zt1qfr1NYHuzeLXfQM9H24wFSUt1Mp8

Track transaction propagation:

thorofare \
  --endpoint1 endpoint1.com:10000 \
  --endpoint2 endpoint2.com:10000 \
  --with-transactions

Full benchmark (slots + accounts + transactions):

thorofare \
  --endpoint1 endpoint1.com:10000 \
  --endpoint2 endpoint2.com:10000 \
  --with-accounts \
  --account-owner 675kPX9MHTjS2zt1qfr1NYHuzeLXfQM9H24wFSUt1Mp8 \
  --with-transactions

Output

Generates JSON with:

• Slot timing comparisons
• P50/P90/P99 percentiles for all metrics
• Account update delays with win/loss counts (if --with-accounts)
• Transaction propagation delays with win/loss counts (if --with-transactions)
• Ping latencies

Delay metrics use 0ms for wins (this endpoint was faster) so the percentiles reflect the full population. A p50 of 0.00 means that endpoint was faster more than half the time.

You can visualize here https://thorofare.triton.one/

Config

See config.toml for gRPC tuning options. Most important:

• tcp_nodelay = true for low latency
• initial_connection_window_size for heavy subscriptions
• initial_connection_window_size for heavy subscriptions (Recommend reading: https://httpwg.org/specs/rfc9113.html#rfc.section.6.9.1, for window size)

Contributing

PRs welcome. Keep it simple:

1. Fork and create a branch
2. Make your changes
3. Test against at least 2 different endpoints
4. Submit PR :D

Related

• UI Repo: https://github.com/rpcpool/yellowstone-thorofare-ui