Faster consensus. Same safety. Smarter performance.
We’re excited to share that Baby Raptr — also known as Optimistic Quorum Store (OptQS) — is coming to the Aptos Network!
This new technique reduces Quorum Store + Consensus latency by two network hops (roughly 150–200ms on Aptos mainnet as of October 2024) in the common case and under all load conditions.
How it works
Today, there are two paths for transactions in consensus:
- For blocks with <100 transactions, Aptos inlines the full transaction payload into the proposal — this gives the best latency (4 hops), but only works when the proposer isn’t bandwidth-constrained.
- For larger blocks, validators rely on proof of store, which requires 6 network hops.
Optimistic Quorum Store introduces a new path — one that offers 4-hop latency even under high load by allowing optimistic batches to be proposed before proof of store is collected.
These batches don’t contain the payload directly but include references. The actual transaction data is disseminated separately via Quorum Store.
What changes?
- Proposers now include both proven and optimistic batches in a proposal.
- Validators vote for a proposal only if they have all batch payloads locally.
- If a payload isn’t available yet, validators try to fetch it. If it doesn’t arrive in time, they timeout the round and move on — just like today.
- A Proposal Status Tracker keeps tabs on which batch authors are slow or missing and helps the proposer make smart choices for future proposals.
The result: throughput stays high and latency stays low, without compromising safety.
What does testing show?
On large-scale Forge testing (150 validators, 15 full nodes):
- ≤100 TPS: No change — inlining and OptQS perform the same.
- ~3000 TPS: OptQS improves latency by 180–200ms.
- ~7000 TPS: Improvements of 200–300ms observed. Further analysis ongoing.
In summary:
4-hop latency path even under high load
150–300ms faster transaction inclusion at scale
Consensus and Quorum Store integrated more efficiently
Safety-first design with adaptive proposer logic
Already showing solid results in testing
Stay tuned as we roll out Baby Raptr and continue to make Aptos faster and smarter for everyone!
Let us know what you think, and feel free to ask questions below