BigQuery to ClickHouse
Three hours of hands-on migration: export a BigQuery public dataset, query it in place, then beat the schema BigQuery hands you three times over.
BigQuery works. That is the problem this workshop is about.
There is a class of product BigQuery structurally cannot serve: customer-facing analytics, in-product dashboards, anything with real users querying concurrently. Its latency floor and slot-based concurrency are architectural, not tuning problems, and because it bills per query, exposing analytics to thousands of customers is exactly where the pricing model breaks.
So this is not a workshop about decommissioning BigQuery. It is a workshop about the dashboard you cannot ship today.
You will export a real BigQuery public dataset, query it in place from object storage before ingesting anything, then take the schema BigQuery's own export format hands you and beat it three times: on compressed bytes, on rows read for a single-user lookup, and on a per-minute conversion funnel that no sort key can rescue.
Learner track
The workshop itself. Start at 00 Setup.
Instructor track
Run-of-show and the floor-helper cheatsheet.
Reference
How to measure without fooling yourself.
What you need
A free ClickHouse Cloud account, created before the session. Nothing installed locally -- every step runs in the SQL console. No Google Cloud project: the dataset is already exported to a public bucket, and the export is demonstrated live so you can reproduce it afterwards at no cost.