Please Implement A Quorum Database
Web page · https://transactional.blog/talk/please-implement-a-quorum-database
First seen by Oriel on .
AI summary
This HPTS'26 talk argues that quorum-based replication is no longer generally impractical, highlighting recent leaderless protocols that improve latency, contention handling, snapshots, and replica catch-up.
- The proposed design criteria are contention-resistant latency, consistent local snapshots, and replica catch-up proportional to missed work rather than total data.
- QuePaxa replaces conflicting Paxos proposers with collaboration, offering a 1RTT fast path and bounded consensus latency of up to 7RTT, but needs throughput optimizations.
- Neo4J's RIOT uses a connected transaction DAG and “Leading Edge” snapshots, reaching 300txn/partition/second and lower latency than its Raft implementation.
- Cassandra's Accord offers a 1RTT fast path and 3RTT worst case, using timestamp-ordered execution for consistent local snapshots and incremental replica catch-up.
- CURP, Jetpack, and Skyros combine 1RTT quorum writes into an unordered operation pool with leadered consensus that later extracts a correct linearizable order.
#quorum replication #leaderless consensus #distributed databases #cassandra
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Points
Score reported by the source.
- Lobsters: 19, latest observation ; +17 over observed period since
Comments
Comment count reported by the source.
- Lobsters: 0, latest observation ; 0 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Lobsters | hgicaq | https://transactional.blog/talk/please-implement-a-quorum-database Read discussion |