CDL
Deep Dive
Google Cloud's serverless data warehouse. Query petabytes with SQL in seconds, with no infrastructure to manage.
BigQuery is a serverless data warehouse — there are no servers to provision, no clusters to configure. You write SQL and Google Cloud automatically scales the compute resources to run it.
The key architecture: storage and compute are completely separated. You can have petabytes of data stored without paying for compute until you run a query.
Columnar architecture
BigQuery stores data in columnar format (Capacitor), not by rows. A query accessing 3 columns of a 100-column table only reads 3/100 of the storage. This makes analytical queries extremely efficient.
Storage/compute separation
Storage in Colossus (Google's distributed file system) is independent of compute in Dremel. You can scale one without the other. It also allows multiple customers to share the same compute.
Standard SQL
BigQuery uses standard ANSI 2011 SQL, not a proprietary dialect. If you know SQL, you can use BigQuery. It supports complex JOINs, window functions, CTEs, arrays, and structs.
Native integration
Connects directly with Cloud Storage, Pub/Sub, Dataflow, Looker Studio, Google Sheets, and dozens of third-party BI tools. Data in Cloud Storage can be queried directly with BigQuery Omni.
On-demand (per query)
Tip: use SELECT specific columns (not SELECT *) and WHERE filters to reduce TB processed.
Reservations (slots)
Enterprise Edition includes committed slots and advanced BI Engine and ML features.
| Capability | What it does | Use case |
|---|---|---|
| BigQuery ML | Train ML models with SQL inside BigQuery. No data export needed. | Regression, classification, time series, recommendations with existing data in BQ |
| BI Engine | In-memory cache to speed up dashboard queries. Sub-second for frequently used tables. | Looker Studio or Looker dashboards with high user concurrency |
| BigQuery Omni | Query data in AWS S3 or Azure Blob Storage from BigQuery with SQL. | Multi-cloud analysis without moving data. Companies with data across multiple clouds. |
| Authorized Views | Share subsets of data without exposing the full table. | Give analysts access to only their columns/rows of sensitive data |
| Streaming inserts | Insert rows with seconds of latency before they are available for queries. | Near-real-time analytics of app events, logs, telemetry |
| Scheduled Queries | Run SQL queries on a schedule to update tables. | Data transformation pipelines without Dataflow or Python code |
| Dimension | BigQuery (GCP) | Redshift (AWS) | Synapse (Azure) |
|---|---|---|---|
| Model | Serverless | Clusters (RA3 = semi-managed) | Serverless + Dedicated pools |
| Query price | $5/TB processed | Per cluster/hour (always on) | $5/TB (serverless) |
| Scale | Automatic and unlimited | Manual node scaling | Auto-scale in serverless |
| ML | BigQuery ML (native SQL) | SageMaker (external) | Azure ML (external) |
| Multi-cloud | Omni: natively AWS + Azure | No | No |
Did you get this topic?
Test what you just learned
A company has 10 petabytes of historical sales data and needs to run complex SQL analyses to generate monthly reports. The data team does not want to manage any server or cluster. Which Google Cloud service is best suited?
Put what you learned to the test with original CDL exam questions.
Start CDL mock examUnlimited mock exams with the Pro plan
Log in to track your progress.