Bigtable (GCP)
A wide-column NoSQL database built for very high throughput at extremely low latency, suited to time-series data, IoT telemetry, and analytics workloads with massive write volume. It is a specialized tool distinct from Firestore's document model or Cloud SQL's relational model.
Frequently Asked Questions
When would you choose Bigtable over Firestore or Cloud SQL?
Bigtable is purpose-built for extremely high write throughput with low, predictable latency at massive scale — think IoT sensor telemetry, time-series metrics, or ad-tech clickstream data with millions of writes per second. Firestore is a better fit for document-style app data with rich querying and offline sync needs; Cloud SQL fits relational data needing joins and transactions. Bigtable trades query flexibility (no joins, no SQL, single-key-based access patterns) for raw throughput and horizontal scale.
What's a common design mistake when modeling data in Bigtable?
Choosing a row key that creates hotspotting — using a sequential timestamp as the row key prefix means all new writes land on the same node range at once, since Bigtable shards by row key range and recent timestamps cluster together. The fix is to reverse or salt the key (e.g., prefix with a hash) to spread writes evenly across tablet servers — a decision that must be made upfront, since changing row key structure later means a full data migration.