GCP Professional Data Engineer Practice Question

Your e-commerce platform must ingest about 200 000 JSON clickstream events per second from users around the world. Engineers need millisecond-level reads on the most recent minute of traffic for real-time personalization, while analysts expect sub-second ad-hoc SQL queries on six months of historical data. You want a fully managed design that minimizes operational overhead and controls storage cost. Which architecture best satisfies both requirements?

Persist events in Firestore (Native mode) and enable automatic export of the collection to BigQuery for analysis.
Stream events into Cloud Bigtable for serving reads; use Dataflow to dump snapshots to Cloud Storage and load them into partitioned BigQuery tables for analytics.
Write events to Memorystore for Redis for low-latency access; export cached keys nightly to BigQuery using Cloud Functions.
Stream events into Cloud SQL (PostgreSQL) via Dataflow; create read replicas and use federated queries for analytics.

Report Issue

Answer Description

Cloud Bigtable is designed for very high write throughput (hundreds of thousands of rows per second) with single-digit millisecond latency, making it well suited for the operational serving workload. A Dataflow pipeline can periodically read from Bigtable and write to date-partitioned BigQuery tables-either by loading files staged in Cloud Storage or by streaming directly-providing a serverless warehouse where analysts can run interactive SQL on months of data at low cost. Cloud SQL cannot ingest or scale to 200 000 writes per second, Memorystore is an in-memory cache without durable storage or analytical interfaces, and Firestore hard-limits writes to roughly 10 000 per second per database, so those alternatives fail the throughput or durability 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.