Sizing Lakebase Search: lessons from a real app and 4 million passages

Databricks published an account of how its engineers sized Lakebase for an internal application called Slide Hub, which indexes roughly 4 million passages. Field engineers use the app to find and reuse content.
The post frames the work as lessons drawn from running a real application at that scale rather than a benchmark exercise. It covers how the team arrived at a database sizing that matched the search workload.
The practical takeaways center on the relationship between the number of indexed passages and the resources the database needs to serve lookups. That connection is the useful part for anyone planning a similar internal search tool.
Why it matters: Teams building search over millions of records can treat this as a sizing reference. The specifics are Databricks' own, so the numbers are starting points for inference, not a universal formula.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Slide Hub is an internal Databricks app. Field engineers use it to find and reuse...