Databricks Lakebase speeds up Postgres database recovery

Databricks just introduced a new way to recover large Postgres databases in its Lakebase managed service. Instead of the old multi-hour restore process, the new architecture lets you create a new database branch at any chosen point in time in seconds—even for databases over 100 TB. This is possible because compute and storage are separate, and restores are a quick metadata action, not a full data copy.

• Traditional restores for large Postgres databases are slow
• Lakebase decouples compute from storage
• Restores now take seconds, not hours
• Restore is a metadata branch, not a full data copy