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 are designing the storage architecture for a transactional PostgreSQL database hosted on an n2-standard-32 (32 vCPUs) Compute Engine VM in a single zone. Your capacity and performance requirements are:
Which storage configuration should you attach to the VM?
Zonal SSD Persistent Disk (pd-ssd) is a durable, solid-state network block storage option in Google Cloud designed for enterprise databases and IOPS-intensive workloads that require sub-millisecond latencies and predictable performance. Unlike ephemeral local storage, data written to pd-ssd is persisted independently of the VM lifecycle.
pd-ssd, IOPS limits are half-duplex, meaning read and write operations share the same IOPS budget. Your peak workload requires 40,000 combined IOPS (28,000 read IOPS + 12,000 write IOPS). A 2,000 GiB pd-ssd volume offers up to 66,000 IOPS (6,000 baseline + 30 IOPS/GiB * 2,000 GiB). Attached to a 32-vCPU n2-standard-32 instance, the per-VM limit caps effective performance at 60,000 IOPS, which comfortably exceeds the 40,000 IOPS target.28,000 * 16 KiB / 1,024) and 187.5 MiB/s of write throughput (12,000 * 16 KiB / 1,024). Because pd-ssd throughput is full-duplex, the 2,000 GiB disk and 32-vCPU VM support 1,200 MiB/s simultaneously for both reads and writes.pd-ssd delivers sub-millisecond latency and preserves database state across VM stops, restarts, and host failures.Compared to a 1,000 GiB pd-ssd (which only delivers 36,000 IOPS and lacks 2-year capacity), NVMe Local SSDs (which lose data on VM stops), and pd-standard (which cannot meet random IOPS or latency targets), a 2,000 GiB Zonal pd-ssd is the smallest standard disk tier that satisfies every performance, capacity, and durability constraint.
Local SSDs are physically attached, solid-state drives located directly inside the physical server hosting the Compute Engine VM. Using the NVMe interface and striping four 375 GiB disks into a 1,500 GiB array delivers ultra-low latency and massive throughput—up to 680,000 read IOPS, 360,000 write IOPS, 2,650 MiB/s read throughput, and 1,400 MiB/s write throughput.
The storage were used strictly for ephemeral workloads—such as scratch processing space, hot read caches, or temporary tablespaces—where data loss on VM restart is acceptable or durability is managed by application-level replication across separate nodes.
While Local SSDs easily surpass the 40,000 IOPS and 625 MiB/s throughput needs of this workload, primary transactional databases requiring durability across VM lifecycle events must use durable block storage such as Persistent Disk or Hyperdisk.
A 1,000 GiB Zonal SSD Persistent Disk (pd-ssd) provides durable, sub-millisecond network block storage whose maximum IOPS and throughput scale linearly with provisioned disk capacity up to the host VM's vCPU limits.
pd-ssd volume provides 36,000 IOPS (6,000 baseline + 30 IOPS/GiB * 1,000 GiB) and 720 MiB/s of throughput. While 720 MiB/s of full-duplex throughput covers the 437.5 MiB/s read and 187.5 MiB/s write rates, pd-ssd IOPS are half-duplex. Because the peak workload demands 40,000 combined IOPS (28,000 reads + 12,000 writes), the 36,000 IOPS ceiling will throttle database queries and increase I/O latency.pd-ssd performance is tied directly to disk size (30 IOPS/GiB and 0.48 MiB/s per GiB above baseline).The database dataset were expected to stay well below 1,000 GiB over the 2-year planning window and peak combined read/write IOPS did not exceed 36,000 IOPS.
A common sizing mistake is comparing peak read IOPS (28,000) and peak write IOPS (12,000) separately against the 36,000 IOPS limit. Because Persistent Disk IOPS are half-duplex, you must sum concurrent reads and writes (40,000 IOPS) before sizing the volume.
Zonal Standard Persistent Disk (pd-standard) is durable network block storage backed by standard hard-disk drives (HDD). It is engineered to provide economical storage capacity and solid sequential throughput for batch and archival workloads rather than high random IOPS.
pd-standard reaches its maximum disk performance—it delivers at most 7,500 read IOPS and 15,000 write IOPS. Because pd-standard IOPS are half-duplex, concurrent reads and writes share this limit, falling drastically short of the 28,000 read IOPS, 12,000 write IOPS, and 40,000 combined IOPS required.pd-standard volume, throughput is half-duplex with caps of 1,200 MiB/s read and 400 MiB/s write, so heavy concurrent reads and writes compete for bandwidth.pd-standard is capped at 7,500 read IOPS and 15,000 write IOPS regardless of disk size.pd-standard throughput requires large sequential I/Os (256 KiB to 1 MiB or larger) and multiple parallel streams.The workload involved large, latency-insensitive datasets (such as data archival, cold logs, or sequential batch analytics) where storage cost per GiB is the primary concern and random IOPS demand is minimal.
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