Blocking database access compared with non-blocking PostgreSQL access.
Tail latency reduction at the highest measured load point.
Throughput improvement at the highest measured load point.
Interactive comparison
Metrics by concurrent users
Average latency
Lower is better. Shows how request duration changes as concurrency grows.
Throughput
Higher is better. Shows completed requests per second.
Tail latency
p95 and p99 are often more important than average latency for real users. Max latency is available only for runs where it was recorded.
Raw benchmark data
Numbers currently reflect the documented JDBC vs pgasync runs in this repository.
| Concurrent users | Scenario | Throughput | Avg latency | p95 | p99 | Max |
|---|