Prepare and test your skills

Prepare and test your skills

Free sample question with answers and detailed explanation below. The Professional Cloud Database Engineer practice bank has an additional 769 practice questions.Try 10 free realistic questions: instant score, no credit card, and no account required.
You manage a self-hosted PostgreSQL OLTP database running on an N2 Compute Engine VM with 4 vCPUs and a 1,000 GiB zonal SSD Persistent Disk (pd-ssd). During peak traffic, queries slow down and Cloud Monitoring reports the following disk metrics:
HealthyAlthough a 1,000 GiB pd-ssd volume supports up to 36,000 IOPS based on capacity, throughput remains well below its limit.
What should you do to resolve this IOPS bottleneck?
Attaching multiple Persistent Disk volumes of the same type (such as two pd-ssd disks) to a single VM combines their capacities when calculating per-disk performance limits, and allows administrators to separate database files (like tablespaces and WAL logs) or stripe volumes at the OS level.
pd-ssd volumes combined.The VM already had sufficient vCPUs to support more than 36,000 IOPS, and you wanted a non-disruptive way to increase total pd-ssd capacity-based IOPS without resizing the original volume.
I/O queue depth controls how many I/O requests are issued in parallel to saturate networked block storage, while Linux readahead (blockdev --setra) pre-fetches contiguous blocks into memory ahead of sequential read requests.
The workload consisted of large sequential scans (256 KiB–1 MiB I/O size) that were falling short of throughput limits due to low concurrency and small OS read sizes, rather than hitting a hard 15,000 IOPS VM cap.
Resizing a Persistent Disk increases its provisioned capacity in GiB, which automatically scales the per-disk IOPS and throughput limits for pd-balanced and pd-ssd volumes (up to the disk type's maximum cap). It can be performed online without detaching the disk or stopping the VM.
pd-ssd volume already provides up to 36,000 IOPS based on disk size, yet Cloud Monitoring shows IOPS plateauing at 15,000 IOPS. Because the 4-vCPU N2 VM caps total attached pd-ssd IOPS at 15,000, doubling the disk size to 2,000 GiB will still result in the exact same 15,000 IOPS ceiling.The VM already had a large vCPU count (such as 16 or 32 vCPUs) whose per-instance IOPS ceiling exceeded the disk's capacity-based limit, and Cloud Monitoring showed the workload saturating the smaller disk's per-GiB IOPS limit.
Changing a Compute Engine VM's machine type to one with more vCPUs scales the compute, memory, and I/O processing capacity allocated to the instance. For Persistent Disk volumes, maximum achievable storage performance is governed by two independent ceilings: the per-disk limit (determined by disk type and provisioned GiB) and the per-instance limit (determined by the VM's machine series and vCPU count).
pd-ssd volumes, even though a 1,000 GiB pd-ssd volume can theoretically deliver 36,000 IOPS (6,000 baseline + 30 IOPS per GiB). Upgrading to an N2 machine type with 8 or more vCPUs raises the per-instance IOPS ceiling.pd-ssd volume without modifying disk capacity.The Healthy status and 8 KiB random I/O size confirm this is an IOPS-bound workload hitting the VM-level vCPU limit, not a disk-capacity limit or hardware degradation. Scaling the VM's vCPUs directly removes the true bottleneck.
Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated Validated