Watch a single transaction travel through Solana's consensus side by side, under the protocol Solana uses today and the one it's moving to.
Under Alpenglow that transaction is final in about 150 milliseconds. Under TowerBFT the same transaction takes about 12.8 seconds. This tool shows you where those seconds go.
Not affiliated with or endorsed by the Solana Foundation yet. "Solana" is a trademark of Solana Foundation, used here only to name the protocol being simulated. The mark above is this project's own, vectorised from
SCL-logo.png.
Built as a free, open-source teaching tool for Solana education programs and for anyone who would rather see consensus than read about it.
Time from the transaction joining a block until it can no longer be undone. TowerBFT reaches confirmed first (optimistic, still reversible), then rooted, which is final; Alpenglow goes straight to final. Seed 1, 25 validators:
| Scenario | Alpenglow → final | TowerBFT → confirmed | TowerBFT → rooted |
|---|---|---|---|
| happy-path | 451 ms (fast path, 80%) | 513 ms | 12.7 s |
| ideal (tx rides the block's last slice) | 150 ms (fast path, 80%) | 187 ms | 12.4 s |
| offline-25pct | 563 ms (slow path, 60%+60%) | 493 ms | 17.8 s |
| leader-down (tx sent into a dead window) | 452 ms after the next live leader | 350 ms | 12.5 s |
| partition-heal (tx sent mid-partition) | 1.3 s after the heal (standstill re-broadcast, then 512 ms of consensus) | 495 ms | 12.7 s |
| twenty-twenty (21% offline, then a partition) | 462 ms (slow path) | 489 ms | not within 14 s |
happy-path and ideal run on the same healthy cluster, yet Alpenglow takes 451 ms in one and 150 ms in the other. The difference isn't consensus. It's when the transaction lands in the block. In happy-path it rides the block's first slice and waits for the leader to finish building; in ideal it rides the last slice, leaving only the voting.
cargo test --workspace
cargo run -p sim-cli -- scenarios
cargo run -p sim-cli -- run --scenario happy-path --protocol both
./scripts/build-wasm.sh # needs: rustup target add wasm32-unknown-unknown, wasm-bindgen-cli 0.2.128
cd web && bun install && bun run dev # http://localhost:5173The CLI also replays saved traces, inspects one validator at a moment in time, and sweeps a parameter into a CSV. Run consensus-lab --help for the rest. Tests: bun run test, bun run e2e. Scenarios are small JSON files where every field has a default; write your own in the app's custom JSON editor.
Click ☰ Lessons. Each loads a scenario and stops the clock at the moment that matters, after asking what you think happens.
- One transaction, two protocols
- Why votes on-chain cost block space
- Skip certificates
- Lockouts and why partitions hurt TowerBFT
- 20+20
- Vote thresholds and certificate types follow Alpenglow white paper v1.1 and SIMD-0326; real BLS signatures aren't simulated: a certificate arrives carrying its stake and is trusted.
- Standstill is modelled on purpose: votes are sent once, so after a split heals each half would think it had already voted. It's set to 2 s here so you can watch; mainnet waits far longer.
- PoH is modelled as the leader's clock rather than hashing, and the network is a table of latencies rather than a packet simulation.
Full list of what is and isn't modelled: docs/architecture.md.
Apache-2.0. Contributions welcome. "Solana" is a trademark of Solana Foundation, used only to name the protocol being simulated.
Further reading: docs/protocols.md both protocols as simulated · docs/architecture.md module map, test map, honest limits · web/README.md render loop, adding a lesson, URL scheme · docs/wasm-api.md engine ↔ UI contract