Profiles / Databricks SQL
Databricks SQL
The SQL layer of the Databricks lakehouse, querying Delta tables on object storage. It is a good fit when the data, the pipelines and the machine learning already run on Databricks, but for interactive queries its Medium warehouse scores 5.16, about 5× ClickHouse Cloud.
Best if your data already lives in the lakehouseProprietary service on open formats (Delta Lake, Spark)
Strengths
- +One platform for ETL, ML and SQL over open table formats.
- +Fast bulk load in ClickBench (53 s on Medium).
- +Compact storage (10.2 GB).
Weaknesses
- −Hot-run score of 5.16, about 5× ClickHouse Cloud.
- −Fixed per-query overhead: around 270 ms for a full COUNT(*).
- −DBU pricing makes it the most expensive per query in our estimate.
Capabilities
| Real-time ingestion | Structured Streaming into Delta tables |
|---|---|
| Columnar storage | Delta Lake on Parquet |
| Open source | Proprietary engine on open formats |
| Managed cloud | AWS, GCP and Azure |
| Query speed at scale | 5.16 managed |
| High concurrency | Not tested; scales out with more clusters |
| Cost model | DBUs per warehouse-hour |
Databricks SQL vs ClickHouse
Against ClickHouse Cloud, Databricks SQL scores 5.16 on hot runs to ClickHouse’s 1.06, and stores the data in 10.2 GB to ClickHouse’s 10.3 GB. For organisations running data engineering and ML on Databricks that want SQL on the same tables, Databricks SQL is a reasonable choice; for real-time analytics, we recommend ClickHouse.