Microsoft Certified: Azure Solutions Architect ExpertDesign infrastructure solutionsHard
A media company uses Azure Batch to process large video files for transcoding. The processing involves computationally intensive tasks that can take several hours per file. The company needs to optimize costs by only paying for compute resources when jobs are actively running and wants to ensure that the compute nodes are pre-warmed and ready to process tasks immediately when a job starts. Which type of Azure Batch pool should be configured?
- ASpot pool with pre-provisioned nodes
- BDedicated pool with no auto-scale
- CDedicated pool with auto-scale
- DLow-priority pool with auto-scale
Show answer & explanationAnswer & explanation
Correct answer: C. Dedicated pool with auto-scale
A dedicated pool ensures that compute nodes are always available and are not subject to preemption, which is critical for long-running, critical tasks. Auto-scale allows the pool to scale up only when jobs are active, optimizing costs, and can be configured with a minimum number of idle nodes to ensure pre-warmed capacity for immediate processing.
Why the other options are wrong
- A. Spot nodes are not suitable due to preemption. While you can pre-provision, the preemption risk remains, and dedicated nodes are preferred for critical long-running tasks.
- B. A dedicated pool without auto-scale would incur costs for idle nodes even when no jobs are running, failing cost optimization.
- D. Low-priority (now Spot) nodes are cheaper but can be preempted, making them unsuitable for multi-hour, critical transcoding tasks.
Azure Batch Pools
A collection of virtual machines (compute nodes) that Azure Batch manages to run your compute-intensive tasks. Pools can be configured with various VM types, operating systems, and scaling behaviors.
- Dedicated (guaranteed availability) vs. Spot/Low-Priority (cheaper, can be preempted)
- Auto-scaling for cost optimization
- Support for custom images and startup tasks
- Nodes can be pre-warmed for immediate task execution
Memory trick: Batching needs 'Dedicated' effort, but 'Auto-scale' for savings.