BigQuery
Serverless, petabyte-scale enterprise data warehouse with built-in ML and BI capabilities.
Serverless, petabyte-scale enterprise data warehouse with built-in ML and BI capabilities.
What is BigQuery?
Google Cloud BigQuery is a fully managed, serverless enterprise data warehouse designed to ingest, store, and analyze petabytes of structured and semi-structured data with blazingly fast SQL execution. Because compute is completely separated from storage via Google's Colossus distributed file system and Jupiter petabit-scale network, you can scale storage infinitely at low cost and dynamically allocate computing power (Dremel execution engine) only when executing queries.
Decision Guide: When to Choose This Service
Use BigQuery for large-scale OLAP SQL queries, analytical dashboards, log analysis, machine learning on tabular data (BigQuery ML), and multi-terabyte data warehousing where zero operational cluster management is desired.
Do NOT use BigQuery as a transactional database (OLTP) requiring low-latency single-row updates/deletes, or as a key-value cache. Pick Cloud SQL, Cloud Spanner, or Firestore instead.
Core Architectural Pillars
Compute slots scale on-demand while storage charges remain flat.
Encodes columns with run-length, dictionary, and vector compression for minimal I/O scans.
Partitioning divides tables by ingestion time, timestamp, or integer range. Clustering co-locates related data based on up to 4 column keys for query pruning.
A slot is a virtual CPU unit evaluating SQL. On-Demand pricing dynamically allocates slots; Editions (Standard/Enterprise) provide dedicated baseline and autoscaling capacity.
Train k-means, linear regression, and XGBoost models directly using standard SQL without exporting data. Omni allows querying data residing in AWS S3 and Azure Blob Storage without egress charges.