Benchmarks

Same workload on both sides, identical settings, every run published with logs and disclaimers. We run the test even when the result wouldn't flatter us.

May 2026

AWS RDS for PostgreSQL vs ServersCamp managed PostgreSQL

pgbench TPC-B (heavy-write OLTP) on three database sizes - in-cache, partial cache, disk-bound. 9 runs total.

Disk-bound (production) ServersCamp wins
+9.0% TPS -48% p95 latency 5× more stable (CoV 4% vs 23%)
In-memory (CPU-bound) AWS wins
-5.8% TPS for ServersCamp AWS uses a newer CPU generation (see disclaimer)
Read benchmark →

On the way

  • The same pgbench run on every major managed Postgres provider - DigitalOcean, Render, Supabase, Neon, Aiven, Crunchy Bridge, and others as we work through them. One workload, identical settings, the same disclaimer template.
  • S3-compatible object storage - latency, throughput, request-rate vs hyperscaler equivalents.
  • VM provisioning end-to-end - API response latency, create-call round-trip, and total wall-clock time from request to SSH-ready, vs hyperscaler general-purpose instances.

Each benchmark documents the exact setup, methodology, every run (not just the median), and what we didn't test. Per-interval logs are linked from each article. If you can't reproduce a result on your account, that's a writeup bug - tell us via a ticket.