Quick Note — Recommend a Compute Solution for Batch Processing
AZ-305 › Unit 4: Design infrastructure solutions › Design compute solutions › Recommend a compute solution for batch processing
Quick Note — Recommend a Compute Solution for Batch Processing
A pharmaceutical research team needs to simulate molecular configurations against a target protein. Each simulation runs minutes on a 4-vCPU box; the entire campaign needs to complete in a weekend. The lead engineer's first design is a VMSS of 50 on-demand VMs running scripts in parallel — accurate, expensive, and operationally heavy. The architect replaces it with Azure Batch using a low-priority pool that scales from zero to 400 nodes, runs the tasks, and scales back to zero on Sunday night. The campaign finishes Saturday afternoon. The bill is one-third of what the VMSS approach would have cost.…
Retrieval target
- Recommend a Compute Solution for Batch Processing
- Closed-book recall
- 5 minutes
- Open the full lesson after a miss
Decision anchors
| Prompt | Compact answer |
|---|---|
Azure Batch | A managed Azure service for large-scale parallel and high-performance computing. Provisions pools of VMs, schedules tasks across them, retries failures, autoscales pool size, and tears down. Pay nothing for the Batch service itself — only for the underlying VM compute. Pick for embarrassingly-parallel finite workloads: rendering, scientific compute, genomics, financial monte-carlo. |
| Batch Pool, Job, Task | Pool — the VM compute fleet (size, SKU, OS image, auto-scale formula). Lives independently of jobs. Job — a logical container for related tasks bound to a pool; defines retry policy, priority, lifetime. Task — a single command executed on one pool node; reads inputs, runs an executable, writes outputs. A pool can run many jobs; a job runs many tasks. Pool sizing and task sizing are tuned independently. |
Batch Low-Priority / Spot nodes | Pool VMs running on unused Azure capacity at a steep discount (~80% off). Microsoft can preempt them when capacity is reclaimed. Batch handles preemption transparently — the preempted task is rescheduled on another node, retried per the job's retry policy. Use for idempotent, restartable workloads — most rendering / HPC fits. |
| Batch auto-scale formula | A simple expression evaluated periodically that returns the target node count based on current state — pending tasks, running tasks, time. Example: $pending = $PendingTasks.GetSample(1); $TargetDedicatedNodes = max(0, $pending / 4); ('1 node per 4 pending tasks'). Tunable interval (default 5 min). Replaces hand-tuned scale rules; you write the math, Batch scales the pool. |
Read the answers once, then cover the right-hand column and reconstruct each one from the prompt. A useful answer names the requirement, the recommended control or service boundary, and the nearest alternative it rejects. If you can only recognize the answer after seeing it, retrieval is not yet secure.
Ninety-second explanation
Without notes, explain:
- What requirement signals this learning objective rather than a neighbouring one?
- Which two solution families are most likely to be compared?
- Which hard constraint eliminates the strongest distractor?
- What identity, network, data, or failure boundary must appear in the design?
- Which operational test would prove the recommendation works?
Then compare your explanation with the full lesson. Record the missing decision rule—not merely the missed product name—in your error log.
When to open the full lesson
Open the curriculum-linked lesson when you cannot explain a comparison, when a scenario depends on a numeric limit or SKU feature, or when the service is on a retirement path. Use current Microsoft Learn documentation for availability, limits, pricing, naming, and migration milestones; the quick note is intentionally compact.
Source and freshness
Derived from the linked AZ-305 lesson and retrieval deck, grounded in both attached course sources. Reviewed 2026-08-02. Current Microsoft documentation controls changing product contracts.