GCP Associate Cloud Engineer Practice Question 84: Local SSD Ephemeral Scratch
Google Cloud Associate Cloud Engineer · Test 9 · question 4 of 10
A batch analytics workload on Compute Engine writes large volumes of temporary intermediate files during processing and discards them once each job finishes. The team wants the lowest possible write latency and highest IOPS for this scratch space, and is fine with losing the data if the VM stops or is terminated. Which disk option best fits this need?
Show answer and explanation
Answer:
Local SSDs are physically attached to the host server, giving very high IOPS and very low latency, but their data does not survive a stop, terminate, or host maintenance event (except live migration in some cases) and they cannot be independently attached to another VM after a hard failure. That tradeoff is exactly acceptable for disposable scratch data and is cheaper and faster than persistent options for this use case.
Option A adds cross-zone replication and durability guarantees that this ephemeral workload does not need, at higher cost and lower peak IOPS than Local SSD. Option B is durable network-attached storage; it survives instance termination, which is unnecessary overhead for data that is intentionally discarded. Option D adds network file system overhead and is meant for shared, durable file storage across many clients, not for the lowest-latency local scratch space of a single instance.
Reference: About Local SSDs