4.5 Knowledge Check - Compute and Elasticity

1.

An architect expects VM Auto Scaling to handle a growing single-VM workload by adding CPU and RAM to the running replicas as load increases. Why is this the wrong mental model of the service?

A VM Auto Scaling is horizontal-only: it adds and removes whole VM replicas under a metric policy and never resizes running ones, and a replica-template change applies only to replicas created afterwards, so growing a single VM is a vertical-scaling job, not an auto-scaling one. B Auto Scaling resizes replicas but only upward, so RAM can grow live while CPU cannot, which is why the workload would stall. C Auto Scaling adds CPU and RAM to every replica at once, but only during the cooldown window, so the change is merely delayed rather than impossible. D Auto Scaling can resize running replicas, but only if the group has a single metric policy, which this design is missing.

2.

An architect is choosing a compute class for a workload that fits neatly into a fixed bundle of vCPU, RAM, and NVMe, but whose persistent business data must survive instance rebuilds. A teammate argues a Cube is unsuitable because "Cubes cannot use Block Storage." How should the architect respond?

A The teammate is correct; choose a Dedicated Core server instead, because only Dedicated Core can attach Block Storage. B The teammate is correct, but it does not matter because the Cube's NVMe volume is itself persistent across rebuilds. C The teammate is mistaken: a Cube can attach add-on Block Storage volumes alongside its NVMe template, so place the persistent data on those volumes, which outlive the immutable template. D Cubes are unsuitable for any production workload because their template properties can change unexpectedly at runtime.

3.

FinCorp needs to place a single-VM database boot-and-data disk on Block Storage and wants the volume to deliver its full rated SSD performance. Which design decision most directly governs whether the SSD will perform as expected?

A Selecting availability Zone 3 for the volume, because SSD performance is only guaranteed in Zone 3. B Choosing HDD instead, because HDD performance scales with volume size while SSD performance does not. C Enabling cross-region replication on the volume so reads are served from the nearest copy. D Sizing the SSD volume at or above the roughly 100 GB floor, because SSD performance scales with volume size and a sub-100 GB SSD runs at suboptimal IOPS.

4.

An architect is tuning a VM Auto Scaling group whose replicas take several minutes to boot and warm up. In testing, the group repeatedly scales out, then immediately scales back in, oscillating under steady load. Which configuration choice best stabilises the group?

A Reduce the cooldown to its minimum so the group reacts faster to each metric reading. B Add a second metric policy so CPU and network are evaluated independently. C Set a cooldown at least as long as the replica warm-up time and keep the scale-in and scale-out thresholds well apart, so new replicas can absorb load before the next action and a single noisy reading cannot trigger opposing actions. D Lower the minimum replica count to zero so the group can fully drain between load spikes.

5.

FinCorp must run its large existing VMware estate under strict EU sovereignty while shifting the platform lifecycle off its own teams, and it also wants new elastic workloads at the edge. Which approach best fits these constraints?

A Re-platform the entire VMware estate onto standard IONOS Dedicated Core auto-scaling tiers to gain elasticity everywhere. B Run the estate on a US-hyperscaler VMware service in a European region, since the region keeps the data in the EU. C Keep the estate on-premises indefinitely, because no managed VMware option can preserve the existing operating model. D Run the regulated estate on IONOS Private Cloud (dedicated VMware, EU-operated control plane, included licensing) and place the new elastic workloads on standard IONOS compute, linked over hybrid connectivity.