Disk cache

Disk cache is the current Databricks name, previously Delta cache (since September 2022).

Delta cache → Disk cache

Automatic caching that accelerates data reads by copying remote Parquet files into a node's local storage in a fast intermediate format, so repeated queries never re-fetch them from cloud storage.

• It carries two former names in a single sentence of its own docs, having been the DBIO cache and then the Delta cache before anyone settled on describing what it actually is. 🔗
• The rename came with an explicit promise of nothing happening: the September 2022 note says the behaviour did not change at all, only the words. 🔗
• To get it you pick the hardware, not a setting - Databricks tells you to choose a worker type with SSD volumes, and those workers arrive with disk caching already enabled. 🔗

⚠ Limitations: On SQL warehouses and Databricks Runtime 14.2 and above the CACHE SELECT command is ignored, and with autoscaling enabled a decommissioned worker takes its cache with it, so the data has to be read from the source again. 🔗

Category

Compute / BI

In use from

September 2022🔗

Also known as

Disk caching, SSD cache

Verified

2026-08-20