Monitor, Troubleshoot, and Optimize Workloads in Azure Databricks DP-750 Practice Questions (Page 4)

Questions 16–20

1. 16expert · hard

   A data platform team manages a large Delta table that is used for both point lookups and full-table scans. The table is partitioned by date and has a high number of small files. The team wants to improve query performance for both access patterns. They have already run OPTIMIZE, but the point lookups are still slow. What should they do next?

   • AEnable change data feed on the table.
   • BIncrease the number of partitions in the table.
   • CUse ZORDER BY on the columns used in point lookups.
   • DRun VACUUM to remove old files.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   A query profile shows that a large amount of data is being shuffled between stages. Which operation is most likely causing this?

   • AA wide transformation like groupBy or join
   • BReading from a Delta table
   • CA narrow transformation like filter or select
   • DWriting to a Delta table
   Reveal Solution Discussion Select an answer first
4. 18expert · hard

   A data engineer is troubleshooting a Spark job that processes a skewed dataset. The DAG shows that one stage has a single task that processes 80% of the data, while the other tasks process very little. The job is running on a cluster with 8 cores. The engineer has already tried increasing the number of shuffle partitions, but the skew persists. What should they do next?

   • AEnable adaptive query execution (AQE) to automatically handle skew.
   • BUse a salting technique to add a random prefix to the join key.
   • CAdd more executor cores to increase parallelism.
   • DIncrease the executor memory to handle the skewed task.
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A team runs multiple interactive clusters in Azure Databricks. They notice that some clusters are underutilized, while others are overutilized. They want to optimize resource utilization and reduce costs. They have already enabled autoscaling on all clusters. What should they do next?

   • AIncrease the autoscaling maximum on all clusters.
   • BUse cluster pools to reduce startup time and share resources.
   • CTerminate all clusters and use job clusters instead.
   • DReduce the autoscaling minimum on all clusters.
   Reveal Solution Discussion Select an answer first
6. 20foundation · easy

   In the Query Profile, which metric indicates the amount of data read from disk versus memory for a given operation?

   • AJoin metrics
   • BScan metrics
   • CShuffle metrics
   • DAggregation metrics
   Reveal Solution Discussion Select an answer first