Databricks Certified Data Engineer - Associate

Explanation

When SQL queries are submitted to a non-running SQL endpoint, there is a cold start time required to provision and start the compute resources. The key issue described is that queries run too slowly specifically when submitted to a non-running endpoint.

Let's analyze each option:

A. Turn on Serverless feature and change Spot Instance Policy to "Reliability Optimized" - This is incorrect because:

• Changing Spot Instance Policy affects cost/availability trade-offs but doesn't directly address cold start times
• The Serverless feature itself is the solution, not the Spot Instance Policy setting

B. Turn on the Auto Stop feature - This is incorrect because:

• Auto Stop automatically stops endpoints after a period of inactivity to save costs
• This would actually increase cold start occurrences, making the problem worse

C. Increase the cluster size of the SQL endpoint - This is incorrect because:

• Larger cluster sizes improve performance for running queries but don't reduce cold start time
• The problem is specifically about queries submitted to a non-running endpoint

D. Turn on the Serverless feature for the SQL endpoint - CORRECT because:

• Serverless SQL endpoints eliminate cold starts by maintaining warm pools of compute resources
• Queries can start executing immediately without waiting for cluster provisioning
• This directly addresses the issue of slow queries when submitted to non-running endpoints

E. Increase the maximum bound of the SQL endpoint's scaling range - This is incorrect because:

• Scaling range adjustments affect how many workers can be added during query execution
• This doesn't help with the initial cold start when the endpoint is not running

Key Concept: Serverless SQL endpoints in Databricks provide instant-on capabilities by keeping compute resources ready in a warm pool, eliminating the need to wait for cluster startup when submitting queries to a non-running endpoint.