When does a PostgreSQL deployment outgrow a single instance?

Three classic triggers: write throughput exceeding the largest instance's WAL bandwidth (~tens of thousands of writes/sec), table sizes where index maintenance vacuum runs longer than the autovacuum interval, or a working set that no longer fits in RAM. Each has a different mitigation — sharding, partitioning, or read replicas.

Is Postgres logical replication production-ready?

Yes, but with caveats. Logical replication is solid for read replicas, zero-downtime major-version upgrades, and CDC pipelines. It still struggles with large transactions, DDL replication, and conflict resolution in multi-writer setups. PG 16+ closed many of these gaps; PG 17 closed more.

Should you use pgvector or a dedicated vector database?

If your application already runs on Postgres and your vector workload is moderate (≲100M vectors, ≲100 QPS), pgvector with the HNSW index is the right answer — same backup, same transactions, same operations team. Above that scale, a purpose-built vector DB starts to pay off.

\u2190 All research

Database

PostgreSQL at scale

Operating PostgreSQL when the dataset and write rate stop fitting on a single instance — partitioning, replication, vacuum tuning, and the managed-service landscape.

PostgreSQL is now the default database for most new applications; the operational complexity of running it at scale used to be a major drawback, but the managed-service ecosystem (Aurora, AlloyDB, Neon, Supabase, RDS) and the extension ecosystem (Citus, TimescaleDB, pgvector) have closed most of that gap.

Notifire's coverage tracks the releases and tooling that change the cost-or-complexity curve: logical replication maturity, partition routing, vacuum behaviour under high write load, and the connection-pooling layer (PgBouncer, pgcat) that sits in front of nearly every production deployment.

PostgreSQL at scale

Latest briefings on PostgreSQL at scale

ClickHouse Unveils Major Product Updates

Critical Drupal Flaw Affects PostgreSQL Sites

New Postgres extension improves data handling

ChatFeatured boosts AI analytics performance

Frequently asked questions

When does a PostgreSQL deployment outgrow a single instance?

Is Postgres logical replication production-ready?

Should you use pgvector or a dedicated vector database?

Related topics