What is all-purpose compute?

All-purpose compute is Databricks compute built for interactive work: notebooks, exploration, debugging. Clusters stay up between commands so people can iterate, and the DBU rate runs roughly 3–4× jobs compute on the same tier. The premium buys interactivity — capability that bills whether or not anyone is using it.

Interactive work needs a cluster that’s already warm when the next command arrives, so all-purpose clusters idle by design. That’s the right trade for a team actively iterating, and pure cost for a cluster nobody has touched since lunch.

Two patterns account for most all-purpose overspend. Generous auto-termination windows: a cluster set to terminate after 120 idle minutes bills two hours of premium-rate DBUs after every session, every day. And scheduled jobs left running on interactive clusters, paying the 3–4× rate for work that belongs on jobs compute.

A useful mental model: the all-purpose rate is rent for immediacy. If a workload doesn’t need a human waiting on the result, it doesn’t need to pay that rent.

What is all-purpose compute?

See these numbers in your own workspaces.