Databricks Certified Data Analyst Associate free question 4 / 10 · Domain 4: Executing queries using Databricks SQL and Databricks SQL Warehouses
A BI team's dashboards are idle most of the night, then get bursts of many queries every morning and after lunch, and users complain about waiting while queries queue. Which SQL warehouse type does the documentation recommend for this pattern, and why?
- Classic, because it adds IO features that speed up selective scans.
- Classic, because a fixed warehouse size avoids startup time entirely.
- Pro, because its compute runs in your account and starts in seconds.
- Serverless, for fast startup and rapid autoscaling under bursts.
Show the answer and explanation
Answer: D
Why this is correct
The warehouse types page says a serverless SQL warehouse supports all the performance features of Databricks SQL. It lists rapid startup, typically between 2 and 6 seconds, rapid upscaling when more compute is needed to keep latency low, and quick downscaling when demand is low. It then advises choosing serverless for the best startup performance, the most efficient IO, smarter handling of query demand that varies greatly over time, and rapid autoscaling when query queuing occurs. Bursty dashboards with queuing match that description, and the quick downscaling keeps the quiet night inexpensive.
Why the other options are wrong
"Classic, because a fixed warehouse size avoids startup time entirely." … The page attributes the best startup performance and the handling of demand that varies greatly over time to serverless, and in its performance table classic supports Photon only. Classic does not avoid startup time.
"Pro, because its compute runs in your account and starts in seconds." … The startup time of typically 2 to 6 seconds is described for serverless SQL warehouses, using the serverless architecture. It is part of the case for that type, not for one whose compute layer is outside it.
"Classic, because it adds IO features that speed up selective scans." … In the performance table, the classic type supports Photon only; Predictive IO, the suite that speeds up selective scans, is not marked for it. That feature list points away from classic for this workload.
Workloads named for serverless
The same page names ETL, business intelligence and exploratory analysis as workloads where a serverless SQL warehouse performs well. The warehouses page also says Databricks recommends serverless SQL warehouses when available, because capacity management, patching and upgrades are handled by Databricks.