HyperRAFT++ is a crash-fault-tolerant (CFT) Raft-family consensus engine. Production consensus runs in SIMULATE mode today. Throughput is not measured on production, so no TPS figure is claimed for the live network. Vote signatures are dual-accept (Ed25519 or ML-DSA-87), and the post-quantum migration is in progress.
Traditional RAFT (100% liveness, O(n) complexity) ↓ + Majority-quorum voting (crash-fault tolerant; does not tolerate lying validators) ↓ + Multi-level Validation (reduces consensus rounds) ↓ HyperRAFT (CFT, multi-level validation; TPS not measured on production)
Understanding the difference between crash-fault tolerance and tolerance of lying validators is critical for enterprise blockchain security:
Tolerates: f < n/2 crash faults (non-malicious) With 7 nodes: tolerates 3 failures Assumption: failed nodes simply stop responding Weakness: cannot handle nodes that lie or send conflicting messages Protocol: Raft, Paxos, Zab (Zookeeper)
Tolerates: f < n/3 lying (malicious) validators With 7 nodes: tolerates 2 lying actors Assumption: failed nodes may actively try to corrupt consensus Strength: safety guaranteed even with adversarial participants Protocol family: Castro–Liskov, Tendermint, HotStuff (reference only)
The 66.66% threshold comes from Lamport, Shostak & Pease (1982) and Castro & Liskov (1999): 2f+1 correct nodes out of 3f+1 total guarantee safety despite f lying nodes. HyperRAFT++ does not use that threshold: it is a CFT Raft engine with majority commit and does not tolerate lying validators. Vote signatures are dual-accept today (ML-DSA-87 or Ed25519), and ML-DSA-87 log-entry signing is implemented but not active in production (node identity shipped, AV11-3792; consensus runs SIMULATE on one node).
Leader node receives transactions ↓ Batches transactions into block candidate ↓ Proposes block to validator network ↓ Validators receive proposal
Validators verify block validity: ├─ Transaction signature verification (ML-DSA-87 or Ed25519) ├─ Gas limit compliance ├─ State consistency checks └─ Smart contract execution simulation ↓ Vote on block acceptance
Leader collects votes: ├─ Need a majority of votes (CFT Raft, n/2 + 1) ├─ Block committed when threshold reached └─ Committed entries are final under the CFT model (no protection against lying validators) ↓ Broadcast committed block to all nodes
HyperRAFT++ is a Raft-family engine and keeps Raft's crash-fault model: it tolerates crash faults — nodes that stop responding. It does not tolerate lying-node faults — nodes that lie, equivocate, or collude. Multi-party deployments that must survive adversarial participants need a protocol that tolerates lying nodes; HyperRAFT++ is not one today.
Classic Castro–Liskov-family protocols have O(n^2) message complexity — every node talks to every other node. HyperRAFT++ uses leader-based proposal (O(n) messages) with majority voting. Throughput is not measured on production, so no speedup over those protocols is claimed. The optimistic fast path skips extra rounds when the network is healthy.
Tendermint locks to 1-second block times and uses Ed25519 signatures. HyperRAFT++ targets 90ms average blocks (not measured on production) via multi-level validation and is moving its consensus layer to ML-DSA-87 (FIPS 204): vote signatures are dual-accept (ML-DSA-87 or Ed25519) today, and ML-DSA-87 log-entry signing is implemented but not active in production (node identity shipped, AV11-3792; consensus runs SIMULATE on one node).
HotStuff achieved linear message complexity but requires 3 rounds per block. HyperRAFT++ reduces to 2 rounds on the fast path and 3 rounds only during view changes. Throughput is not measured on production, so no throughput multiple is claimed.
No TPS figure is published for the production network. Lab runs are not production measurements and are not quoted here.
10% network load: 90ms block time 50% network load: 120ms block time 90% network load: 180ms block time (degraded but operating) 100% network load: Queuing, maintains integrity
Full Validators (n = 100)
Archive Validators (n = unlimited)
Light Validators (n = unlimited)
DUAL_ACCEPT
# Install Aurigraph node aurigraph install --validator # Generate keypair aurigraph keygen --algorithm=ml-dsa-87 # Bond tokens to become validator aurigraph stake --amount=1000000 --validator # Sync blockchain and join consensus aurigraph run --mode=validator
# View validator set aurigraph status --validators # Monitor consensus rounds aurigraph monitor consensus # Check block finality aurigraph query finality --block=12345
Explore: About · Products · Features · Whitepapers · Sustainability · Tokenized carbon credits · Real-world asset tokenization · SDK · Blog · Plain-text site index · Sitemap · Contact: contact@aurigraph.io