All tools
Free architecture tool

GCP Database Selector

Cloud SQL, AlloyDB, Spanner, Firestore, Bigtable or BigQuery? Answer six quick questions and get a senior-level recommendation — with the reasoning, the alternatives and when NOT to use it.

No sign-up. Nothing leaves your browser. Instant answer.

Questions answered: 0 / 6
  1. 1What shape is your data?
  2. 2How is it accessed?
  3. 3What scale do you need?
  4. 4Consistency needs?
  5. 5Existing engine?
  6. 6Operational model?

How I decide in the real world

  • Pick the access pattern first — OLTP, analytics, cache or real-time. The workload chooses the engine, not the other way around.
  • Don’t reach for Spanner until you truly need horizontal scale or multi-region strong consistency; Cloud SQL covers most OLTP.
  • Keep transactional and analytical data apart — serve from Cloud SQL/AlloyDB, analyse in BigQuery, and stream between them.
  • Match the operational model to the team — the best database is the one your team can actually run at 3 a.m.

Common questions

Cloud SQL or AlloyDB?

Cloud SQL is the right default for most relational OLTP workloads. Reach for AlloyDB when a PostgreSQL workload gets heavy or needs analytical queries next to transactional ones — it keeps Postgres compatibility but adds real performance headroom via its columnar engine.

When should I use Spanner instead of Cloud SQL?

Spanner earns its keep when a single primary hits its vertical limits, or when you need strong consistency across multiple regions with horizontal scale. For a normal app that fits Cloud SQL, Spanner means paying for capabilities you are not using.

Firestore or Bigtable?

Firestore is a serverless document database with real-time sync — ideal for app back-ends that scale to zero. Bigtable is wide-column NoSQL for massive throughput (time-series, IoT, metrics) at petabyte scale. Use Firestore for application data, Bigtable for relentless high-throughput writes.

What is BigQuery for?

BigQuery is a serverless data warehouse for petabyte-scale SQL analytics — not OLTP. Transactional data lands in BigQuery via streaming or CDC, and analysts query it there. Serve app traffic from Cloud SQL or Firestore, analyse in BigQuery.

Guidance based on 2026 Google Cloud capabilities and real-world platform experience — not a substitute for a full architecture review. Uses no cookies and sends no data server-side.