GCP Professional Cloud Architect Practice Question

Your retail analytics platform currently persists clickstream events in a 20-node self-managed MySQL cluster on Compute Engine. Projections show ingest volume increasing ten-fold within 12 months, reaching several terabytes per day and ultimately petabyte scale. The workload is append-only, requires single-digit-millisecond writes and high read throughput by device-ID and time range, and does not need cross-row transactions. You want to minimize operational overhead while ensuring seamless horizontal growth. Which redesign best meets these requirements?

Deploy multiple zonal Compute Engine instances each with local SSDs fronted by a Regional TCP Load Balancer, expanding nodes as data grows.
Replace the cluster with Cloud SQL for MySQL, enable automatic storage increases, and add five read replicas across regions.
Migrate to Cloud Spanner with a multi-region instance and use interleaved tables for efficient range queries.
Store events in Cloud Bigtable using a composite row key of device ID and reversed timestamp, and enable TTL-based garbage collection.

Report Issue

Answer Description

Cloud Bigtable is a fully managed, wide-column NoSQL service that automatically scales by adding nodes and can handle petabytes of data with consistent low-latency reads and writes. A composite row key such as device-ID#reverse-timestamp keeps related events together and supports efficient range scans, while TTL-based garbage-collection policies automatically remove aged data. Cloud Spanner would work but adds unnecessary relational and transactional features at higher cost; Cloud SQL is limited in capacity and would require manual sharding; relying on local SSDs behind additional VMs increases operational burden and does not provide effortless horizontal scaling.

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.