Question 5

Your team runs dozens of short Lakeflow Jobs on classic job clusters throughout the business day. Engineers complain that each job spends several minutes acquiring VMs from the cloud provider before any Spark work begins, and the total cluster start time dominates the short jobs. Serverless compute is not available for these particular JAR-based tasks, so you must use classic compute. You want to reduce cluster start and autoscaling times for these classic clusters without paying Databricks Units (DBUs) for capacity that is sitting idle and ready. Which approach should you implement?

• AIncrease each job cluster's `min_workers` so the cluster never has to scale up from a cold start.
• BAttach the job clusters to an Azure Databricks pool that keeps a set of idle, ready-to-use instances on standby.
• CDisable auto-termination on the job clusters so they stay warm between runs.
• DSwitch the clusters to a larger driver node type so VM acquisition completes faster.
• EEnable Photon acceleration, which pre-provisions worker VMs before the job starts.