Compute Service Tradeoffs and Selection: 10 practice questions

Question 1 of 10

Which compute family is on-demand, automatically managed, and scaled to workload requirements?

1. ASQL warehouses limited to classic deployment and manual scaling
2. BServerless compute with on-demand, automatically managed scaling
3. CDedicated compute shared automatically across all workspace users
4. DClassic compute with provisioned, user-managed resources

Question 2 of 10

Which compute option is optimized for SQL queries, analytics, and business intelligence?

1. AA SQL warehouse optimized for SQL queries, analytics, and BI
2. BServerless job compute optimized as the execution engine for BI queries
3. CStandard compute assigned to a single user or group for analytics
4. DDedicated compute shared among multiple users for SQL querying

Question 3 of 10

Which items make up total compute cost for non-serverless resources?

1. ADBUs plus virtual machine, disk, and associated network costs
2. BDisk and network charges alone, with compute usage excluded
3. CDBUs alone, with disk and network included for every compute type
4. DVirtual machine costs alone, with DBUs charged only for storage

Question 4 of 10

A workload needs a compute resource assigned to one user or group. Which type fits?

1. AA SQL warehouse restricted to one user by definition
2. BDedicated compute assigned to the specified user or group
3. CServerless compute assigned permanently to one named group
4. DStandard compute shared among collaborating users with isolation

Question 5 of 10

Which combination is unsupported on serverless compute?

1. AR workloads and Spark RDD APIs
2. BPython and Spark Connect APIs
3. CSQL notebooks and automatic infrastructure management
4. DOn-demand scaling and query-profile inspection

Question 6 of 10

A job must run longer than seven days. Which action respects the serverless runtime limit?

1. ASet an eight-day query timeout to override the maximum runtime
2. BBreak it into smaller runs or move it to classic compute
3. CKeep one serverless run; no maximum applies
4. DKeep one serverless run and rely on a retry after termination

Question 7 of 10

How are idle instances in a pool billed?

1. AThey incur no DBUs, but cloud-provider billing still applies
2. BThey incur both charges at the active-compute usage rate
3. CThey incur DBUs, but cloud-provider billing stops while idle
4. DThey incur neither DBUs nor cloud-provider infrastructure charges

Question 8 of 10

A BI team runs interactive SQL with bursts of concurrent dashboard queries and wants capacity to adjust to demand without staff managing infrastructure. Which option best fits?

1. AClassic dedicated compute assigned to one dashboard developer
2. BA serverless SQL warehouse for optimized analytics and automatic scaling
3. CClassic standard compute for shared notebooks and user-managed resources
4. DServerless job compute for non-interactive Lakeflow Jobs execution

Question 9 of 10

A scheduled ETL workload is non-interactive, needs isolation between runs, and should avoid paying startup separately for every task. Which choice minimizes cost while meeting those constraints?

1. AA serverless SQL warehouse used as the engine for each ETL task
2. BInteractive classic compute restarted separately for every task in the job
3. CAll-purpose compute kept active so every scheduled run shares one instance
4. DJob compute with a new instance per job and reuse across its tasks

Question 10 of 10

A team wants shared multi-user resources, but its workload requires both GPU-enabled compute and Databricks Runtime for ML. Which choice satisfies the technical requirements?

1. AStandard compute because Databricks Runtime for ML supports this access mode
2. BStandard compute because both restrictions apply only to dedicated resources
3. CDedicated compute because the required features disqualify standard compute
4. DStandard compute with shared resources because that access mode supports GPUs