Databricks SQL

Databricks SQL

Databricks SQL Warehouses on the Lakehouse

Databricks 2,717 SQL runs warehouse-style SQL directly on Delta Lake 237,929 or Iceberg 129 tables in your own object storage, the lakehouse pattern of Lakehouses, Data Quality and Governance, with the vectorized C++ Photon engine. A SQL warehouse comes in three types: serverless (compute in Databricks' account, starting in typically 2 to 6 seconds, with intelligent workload management), pro and classic (compute in your cloud account, about 4 minutes to start). Sizes run from 2X-Small upward, and clusters are added for concurrency. Tables are ordinary Delta tables; liquid clustering replaces fixed partitions as the pruning layout:

A Delta fact table with liquid clustering in Databricks SQL (not run here)SQL

CREATE TABLE booknest.fact_sales (
  order_date DATE, genre STRING, country STRING, qty INT, gross_amount DECIMAL(10,2))
CLUSTER BY (order_date, genre);        -- OPTIMIZE reclusters files incrementally

Billing is in Databricks Units (DBUs) per hour of warehouse time. Azure 6 's price list for East US shows $0.70 per DBU for serverless SQL, which includes the compute, $0.55 for pro and $0.22 for classic, where the virtual machines are billed separately by the cloud provider. Price per DBU differs by cloud, region and tier, so compare the DBUs a size consumes, shown in the warehouse settings, multiplied by the rate.

MENU