BigQuery
BigQuery is Google Cloud’s fully managed, serverless data platform for SQL analytics, ML, and BI at petabyte scale. Storage and compute scale independently with no cluster sizing. We assess it when GCP is in play or a columnar EDW beats stretching Postgres for analytics volume.
Blurb
BigQuery is a fully managed, AI-ready data platform that helps you manage and analyze your data with built-in features like machine learning, search, geospatial analysis, and business intelligence.
Summary
What it is: Columnar warehouse with standard SQL, partitioned tables, streaming inserts, BigQuery ML, and native ties to Looker, Vertex AI, and GCP IAM. Pay per query and storage; slot reservations for steady workloads.
When to use: Large analytics on GCP; federated queries over GCS objects; marketing and product event pipelines landing in tables; teams already standardized on dbt-core against BigQuery.
When to skip: Primary OLTP (use Postgres). Multi-cloud strategy that forbids GCP-only warehouses. Small datasets where Postgres replicas plus Grafana or Metabase suffice.
Key features: Serverless scaling, nested and repeated fields, sharing and authorized views, scheduled queries, cost controls via quotas and reservations.
Details
| Topic | Notes |
|---|---|
| Deploy | GCP project and dataset IAM; no VMs to patch |
| Cost | Monitor bytes scanned; partition and cluster tables; use cached results |
| BI | First-class in Metabase, Redash, Apache Superset |
Practices: Separate dev and prod projects; avoid SELECT * on wide tables; load via batch or streaming with clear retention; pair with Postgres for transactional source of truth.
References