PROVIDER Β· Serverless cloud data warehouse
Google BigQuery
Google BigQuery pioneered the serverless cloud warehouse, billing scanned bytes so petabyte analysis needs no capacity planning, with partitioning, clustering, and BI Engine governing both bills and responsiveness.
Pros
Cons
Product Information
Detailed Notes
Overview
Google BigQuery became generally available in 2011 as Google Cloud's serverless data warehouse out of the internal Dremel paper, fully managing storage, compute, and networking so users only write SQL, and pioneering billing by bytes scanned.
For Backend & Database work the advantage is zero operations with extreme elasticity, since aggregate queries over hundreds of millions of rows return in seconds, bursty analysis needs no pre-scaling, partitioned and clustered tables cut scanned volume, BI Engine accelerates frequent dashboards in memory, and BigLake queries data in object storage directly.
Cost hinges on scanned volume, because billing follows bytes read at about USD 6.25 per terabyte and an unpartitioned wide table scanned fully produces a large bill, so steady workloads should move to slot reservations with partition pruning and result caching.
Core Strengths
- Fully serverless with no capacity planning for petabyte-scale queries
- Partitioning and clustering cut scanned bytes and therefore the bill
- BI Engine accelerates dashboards to sub-second responses in memory
- BigLake unifies object storage and Iceberg tables, reducing data movement
Product Ecosystem
On-demand and capacity pricing
On-demand bills scanned bytes while capacity mode buys slot commitments for steady heavy load.
Partitioning, clustering, and materialised views
Date or enum partitioning with clustered columns is the first cost lever, and materialised views refresh incrementally.
BigLake and external tables
Query Parquet, Iceberg, and Delta data in GCS alongside Dataproc and Dataflow for a complete data platform.
Drawbacks
- Byte-scanned billing makes full scans of unpartitioned tables expensive
- Not directly available in some markets, requiring data residency review
- Slot reservations need capacity planning skill with idle waste if oversized
- Cross-cloud interop with AWS or Azure adds movement and egress cost
Use Cases
- Ad hoc analysis and exploration (β β β β β ) - Serverless on-demand scaling
- Enterprise BI reporting (β β β β β ) - Sub-second responses with BI Engine
- Querying directly over the lake (β β β β β ) - BigLake without movement
- Frequent small OLTP queries (β ) - Scanned-byte billing is poor value
Pricing
| Plan | Price | Scope |
|---|---|---|
| On-demand | About USD 6.25 per TB scanned | First 1 TB monthly free |
| Capacity (slots) | Per slot-hour or commitment | Better for steady load |
| Storage | From about USD 0.02 per GB-month | Active and long-term rates differ |
Cost control centres on scanning fewer bytes through time partitioning, clustered columns, column pruning, result caching, and materialised views, with slot commitments cheaper once monthly scans pass tens of terabytes steadily, and custom quotas essential on development projects.
FAQ
-
BigQuery or Snowflake? BigQuery for GCP investment with serverless elasticity and direct lake querying, Snowflake for multi-cloud needs with precise compute isolation and cross-org sharing.
-
Why do bills spike suddenly? Most often a full scan of an unpartitioned table or an exploratory query without limits, followed by BI tools refreshing unaccelerated dashboards frequently.
-
Are slot reservations worth it? Yes when monthly scan volume is steady and predictable, typically saving over 30 percent, while variable workloads stay flexible on demand pricing.