Google BigQuery — Serverless Data Warehouse

BigQuery architektura, partitioning, clustering, ML a cost control.

Architektura

Oddělený storage (Colossus) a compute (Dremel). On-demand $5/TB nebo flat-rate slots.

Partitioning a Clustering

CREATE TABLE dataset.events
PARTITION BY DATE(event_timestamp)
CLUSTER BY user_id, event_type
AS SELECT * FROM dataset.raw_events;

Dramatické snížení skenovaných dat = nižší cost.

BigQuery ML

CREATE OR REPLACE MODEL dataset.churn_model
OPTIONS(model_type='LOGISTIC_REG', input_label_cols=['churned'])
AS SELECT days_since_last_login, total_purchases, churned
FROM dataset.user_features;

Shrnutí

BigQuery = nejrychlejší cesta k analytice petabytů. Partitioning + clustering = klíč k nákladům.