Citus benchmark shows high efficiency compared to PlanetScale Neki
Key Takeaway
StackGres benchmarks show open-source Citus matching PlanetScale's Neki performance using significantly fewer data nodes and vCPUs.
Summary
A new benchmark from StackGres compares the open-source Citus sharding solution against PlanetScale's Neki. Using a new Query Router architecture in StackGres 1.19 to remove coordinator bottlenecks, Citus matched Neki's 1M and 10M QPS benchmarks while utilizing roughly half the data nodes and significantly less total vCPU.
Technical details of the Citus vs Neki benchmark
- Citus achieved 10M QPS using 24 workers and 48 query routers (3,072 total cores).
- At the 10M QPS mark, Citus used 48% of the data nodes required by Neki.
- Neki required 1.5 to 1.8 times more vCPU to reach comparable performance levels.
- The benchmark utilized PostgreSQL 18.4, Citus 14.1.0, and Graviton4 (arm64) hardware on AWS EKS.
- Workload consisted of single-row primary-key reads using pgbench scripts.
Why it matters
The results suggest that open-source sharding solutions like Citus can achieve massive scale with higher hardware efficiency than proprietary alternatives.
Original reference: postgr.es →