GCP Professional Cloud Architect Practice Question

You are migrating a latency-sensitive pricing engine to Compute Engine. Each request runs mostly single-threaded code and must complete in under 5 ms. The process holds a 1.8 TB in-memory cache and requires at least 250 000 random 4-KB read IOPS at sub-millisecond latency, while write traffic is minimal. The dataset can be rebuilt from Cloud Storage within 10 minutes, so storage does not need to survive a VM restart. Which combination of machine type and block storage will satisfy the performance target at the lowest cost?

A single c3-highcpu-88 VM and a 2 TB Extreme Persistent Disk provisioned for 160 000 IOPS
A single c3-highmem-60 VM with 2.4 TB of Local SSD striped as one volume
Two n2d-standard-64 VMs, each using a 1 TB Hyperdisk Balanced volume
A single m3-ultramem-160 VM backed by a 2 TB Balanced Persistent Disk

Report Issue

Answer Description

The workload benefits from high single-thread performance, a large memory footprint, and very high random-read IOPS with sub-millisecond latency. Compute-optimized C3 VMs deliver Google's highest per-core performance, and the high-mem variant provides sufficient RAM for the 1.8 TB cache. Only Local SSD offers well above 250 000 random read IOPS with <1 ms latency; Extreme Persistent Disk is limited to about 120 000 IOPS per instance, while Balanced or Hyperdisk Balanced volumes fall far short. Because the data can be regenerated, Local SSD's lack of persistence is acceptable and avoids the additional cost of high-IOPS persistent disks. Therefore, a c3-highmem VM with roughly 2.4 TB of attached Local SSD is the most cost-effective solution. The other options either cannot meet the IOPS target or pay for unnecessarily expensive persistent storage that the workload does not require.

Ask Bash

Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.