Postgres isn't slow, your storage is

A presentation at POSETTE 2026 examined how storage performance, rather than PostgreSQL itself, causes common scaling issues like slow ingestion, inconsistent read latency, autovacuum lag, and replication delays when workloads reach hundreds of gigabytes or terabytes. The speaker benchmarked the same 3.3-billion-row PostgreSQL workload on baseline gp3 EBS versus local NVMe storage and proposed production architecture patterns using local NVMe to address these performance bottlenecks. The key insight is that as data exceeds available memory, physical I/O demands increase, and slower storage creates queuing and latency that affects both user queries and background maintenance operations.