Postgres at scale: what breaks (and what doesn't) when live data grows

Abstract

Postgres runs everything until it doesn’t. Time-series and high-ingest workloads eventually hit the same wall: index bloat, vacuum overhead, and a query planner that stops cooperating once tables cross a certain size. Most teams’ next move is to rip Postgres out and stand up something new, which is a decision that trades one set of problems for another.

In this session, Tiger Data’s Matty Stratton walked through exactly where vanilla Postgres starts to strain under scale, what changes structurally when you extend it with TimescaleDB, and via a live demo what actually breaks at scale versus what doesn’t. Plus, you’ll leave with a concrete scorecard for your own systems.

What you'll learn

• Where and why vanilla Postgres degrades under time-series / high-ingest load i.e. the specific bottlenecks, not just "it gets slow"

• What changes structurally with TimescaleDB/Tiger Cloud (hypertables, compression, continuous aggregates) and why it avoids a rip-and-replace migration

• A concrete "what breaks / what doesn't" scorecard attendees can hold their own systems against

• When not to make the switch i.e. what are the honest limits

• Why AI agent workloads are increasingly one of the things pushing teams into this scaling wall in the first place