GCP Professional Data Engineer Practice Question

A media-streaming company runs a multi-region GKE web application. Each request makes metadata lookups that must finish in under 2 ms; 80 % of those lookups repeat within ten minutes, and the Cloud SQL backing store is saturating despite read replicas. The team wants a fully managed in-memory layer that can deliver sub-millisecond reads of JSON blobs (<50 KB), provide automatic failover transparent to clients, include Google-handled maintenance, and evict least-recently-used items when memory is constrained. Which service best meets these needs with minimal operations effort?

Memorystore for Redis in Standard Tier
A regional Cloud Storage bucket configured with Object Lifecycle Management
Cloud Spanner with read-only replicas
Cloud Bigtable with a multi-cluster instance

Report Issue

Answer Description

Memorystore for Redis is a fully managed, in-memory data store that consistently delivers sub-millisecond latency for objects well under 1 MB. The Standard Tier provides a primary-replica architecture with automatic detection of node failure and transparent failover, so clients do not need custom logic. Google manages patching, monitoring, and scaling, reducing operational overhead. Redis supports built-in eviction policies such as Least Recently Used (LRU), allowing the cache to shed old items when it reaches its memory limit. Bigtable and Spanner are low-latency, but they are disk-backed and cannot match Redis cache performance for <2 ms lookups. Cloud Storage is object storage with millisecond-to-second latency and no automatic eviction logic. Therefore, Memorystore for Redis is the only option that meets all stated requirements.

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.