🚀 Leios Demo — pipelined Ouroboros consensus (visualizer)
Runs the official IOG sim-rs 1.4.0 Leios simulator on adatool's backend. Block structure: Input Blocks (IB, data availability) + Endorser Blocks (EB, aggregation) + Ranking Blocks (RB, current Praos-style ordering). Leios claims 10-65× throughput uplift over Praos. Status (2026-04): not on mainnet. 2025 IO Engineering Core funded research ($7.10M); 2026 IO Consensus ($6.65M) targets mainnet-ready implementation.
⚡ Live simulator — IOG ouroboros-leios sim-rs 1.4.0 (real protocol)
adatool.net's backend runs the official IOG Rust simulator on-demand. Each run takes ~1-5s and produces real protocol events — not a mock. Source: github.com/input-output-hk/ouroboros-leios · sim-rs 1.4.0 tag.
Preset:Slots:
Click "Run simulation" to execute an actual Leios protocol simulation. Run multiple presets to compare side-by-side.
📊 All scenarios side-by-side (SPO resource impact comparison)
⚠ Note: "actual" = measured on-chain (current 20.7% block fill). "max" = theoretical full-capacity assuming demand fills the available block space. Real Leios TPS depends on demand — the "max" figures are upper bounds, not guaranteed.
| Metric | Praos (actual) measured | Praos (max fill) theoretical max | Leios @ current demand measured | Leios 10× TPS theoretical max | Leios 32× TPS theoretical max | Leios 65× TPS theoretical max |
|---|---|---|---|---|---|---|
| TPS (effective) | 0.33 | 1.59 | 0.33 | 3.30 | 10.56 | 21.45 |
| × Praos current | 1.0× | 4.8× | 1.0× | 10.0× | 32.0× | 65.0× |
| Bandwidth | 0.5 KB/s | 2.3 KB/s | 0.5 KB/s | 4.7 KB/s | 15.1 KB/s | 30.6 KB/s |
| Storage / month | 1.3 GB | 6.0 GB | 1.3 GB | 12.5 GB | 40.0 GB | 81.3 GB |
| Storage / year | 15 GB | 72 GB | 15 GB | 150 GB | 480 GB | 975 GB |
| 1TB SPO disk lasts | 66.7 yr | 13.8 yr | 66.7 yr | 6.7 yr | 2.1 yr | 1.0 yr |
| CPU (peak / rec.) | 0.6 c spec: 4 c | 1.4 c spec: 3 c | 1.0 c spec: 4 c | 5.6 c spec: 11 c | 8.1 c spec: 16 c | 9.6 c spec: 19 c |
| RAM (peak / rec.) | 5.3 GB spec: 24 GB | 7.2 GB spec: 14 GB | 6.8 GB spec: 24 GB | 35.2 GB spec: 70 GB | 50.3 GB spec: 101 GB | 59.5 GB spec: 119 GB |
| SPO barrier* | 🟠 Medium | 🟠 Medium | 🟠 Medium | 🔴 High | 🔴 High | 🔴 High |
| 💰 AWS EC2 月額 (est) | $325 m6i.2xlarge (8c/32GB) | $185 m6i.xlarge (4c/16GB) | $325 m6i.2xlarge (8c/32GB) | $797 r6i.4xlarge (16c/128GB) | $1616 r6i.8xlarge (32c/256GB) | $1735 r6i.8xlarge (32c/256GB) |
| 💰 GCP 月額 (est) | $360 n2-standard-8 (8c/32GB) | $225 n2-standard-4 (4c/16GB) | $360 n2-standard-8 (8c/32GB) | $836 n2-highmem-16 (16c/128GB) | $1724 n2-highmem-32 (32c/256GB) | $1976 n2-highmem-32 (32c/256GB) |
| 💰 自前サーバー (amortized/mo) | $65 Ryzen 5 / 32GB / 2TB | $65 Ryzen 5 / 32GB / 2TB | $65 Ryzen 5 / 32GB / 2TB | $110 Ryzen 9 / 64GB / 4TB | $205 Threadripper / 128GB / 8TB | $205 Threadripper / 128GB / 8TB |
⚠ これらは全て推定値です — 実機での確認を強く推奨
「peak」= 実際のノードRSS/CPU使用量の推定値 (私たちの mainnet ノード実測 5.3 GB / 0.6 core をベースに log2 スケーリング)。「rec. (recommended)」= 安全マージン込みの推奨スペック (Praos: IOHK SPO ガイド 4 cores / 24 GB; Leios: Galois/IOHK simulation 目安で peak の約2倍)。 推奨スペックは bulk sync 時のスパイク、reorg 時のメモリ増、mempool 急増、ノード + db-sync + monitoring の同居を見込んだもので、平常時の実使用量より大きい。「Leios @ current demand」は Leios 有効化しても需要が現状 (0.33 TPS) のままの場合 — プロトコルオーバーヘッド (IB propagation, voting, EB tracking) で若干増えるが、TPS もストレージも変わらない。 ホスティングコストは standard 価格ベースの ballpark で、リザーブインスタンス/ボリューム割引で大きく変動。本番判断には必ず自身の testnet / staging で実測してください。
* SPO barrier estimate: 🟢 Low = home-run feasible (≤ 12 GB RAM, ≥ 5 yr disk), 🟠 Medium = small VPS feasible, 🔴 High = may require enterprise-grade hardware. Baselines (Praos) measured on our own mainnet cardano-node 10.5.1: RSS ~5.3 GB, CPU ~0.6% steady-state; block data growth ~1.25 GB/month from adatool db-sync archive (last 6 months). Leios projections use log2 scaling calibrated to IOHK/Galois simulation recommendations. Storage assumes demand fills capacity (actual may be less if demand lags). Hosting cost estimates: AWS on-demand Graviton / Intel pricing, GCP standard E2/N2 pricing, self-hosted amortized over 3 years.
How Leios works (simplified)
1. Input Blocks (IB)
Data availability layer. Produced by stake-weighted SPO election at high rate (multiple per slot). Contains actual transactions. IBs are cheap to produce — they don't need to be canonically ordered yet.
2. Endorser Blocks (EB)
Aggregation layer. EB producers vote on which IBs from recent slots to include. Conceptually compresses "which IBs are valid" into a single reference that RBs can point to.
3. Ranking Blocks (RB)
Ordering layer. Current Praos block, but now references an EB (which references IBs). Settles transactions from across multiple IBs at once. Slot schedule unchanged from Praos.
⚠ Honest caveats (why actual gain may be less than theoretical)
• Current mainnet block fill is only 20.7% — before Leios, Cardano has ~5× unused capacity by simply filling existing blocks. Leios unlocks more capacity, but real TPS also depends on demand to fill it.
• Storage: measured from adatool's db-sync archive, current Cardano L1 block data growth is ~1.0-1.5 GB/month (~12-18 GB/year) averaged over the last 6 months. At Leios 65× linear scaling, this becomes ~845 GB/year — a small SPO 1TB disk would fill in ~1.2 years IF demand fully uses the 65× capacity. The 2026 IO Consensus proposal does NOT publish an SPO storage forecast or upgrade subsidy plan.
• CPU + RAM: measured on our own cardano-node 10.5.1 running mainnet: RSS ~5.3 GB, CPU ~0.6% steady-state (db-sync adds another ~13 GB RSS, ~57% CPU). Leios adds IB validation + in-flight buffers + peer state. Galois/IOHK simulation (2024) suggests 10× Leios needs ~8 cores and 48-64 GB RAM recommended; 65× may need 16+ cores and 64-128 GB. Low-spec VPS saw timeouts at 65× in red-team simulations. The proposal does NOT publish formal SPO resource recommendations.
• Bandwidth: each SPO must still download all IBs to verify/vote. High IB rates increase bandwidth demand proportionally, which may exclude home-run SPOs.
• Demand: the 10-65× claim assumes Cardano sees enough mempool demand to fill Leios capacity. Current mainnet has never sustained this level of transaction volume outside of the 2022 NFT mint rush.
Status (2026-04): Leios research implementation exists in a cardano-node research fork. Testnet runs have demonstrated the IB/EB/RB structure. Not yet on public testnet or mainnet. 2025 funded design + simulation + preliminary implementation ($7.10M). 2026 IO Consensus proposal ($6.65M) targets mainnet-ready code, conformance tests, SPO workshops, red-team testing.
Paper: iohk.io/en/research/library — "Ouroboros Leios: Design Goals and Concepts" and subsequent iteration papers.
This demo is a mock: timings are accelerated (1 slot/sec here; real Cardano is 1 slot = 1 sec already), and no actual Leios implementation is running — the block stream visualization and TPS/storage math are illustrative of the protocol's parameters and trade-offs.