GCP Professional Data Engineer Practice Question

Your manufacturing company operates 50,000 IoT sensors that emit one reading every 100 ms, resulting in several terabytes of data per day. You must build a managed storage layer that ingests this stream continuously, keeps 30 days of history, and powers real-time dashboards that query recent readings for individual devices with single-digit-millisecond latency. The schema consists of timestamp, device_id, and numeric metrics that evolve over time. Which Google Cloud service and design best satisfy these requirements while minimizing operational overhead?

Store each reading as a document in Firestore in Datastore mode, keyed by device_id/timestamp, and query with composite indexes.
Ingest the sensor stream into Cloud Bigtable and design the row key as reverseTimestamp#deviceId to support low-latency range scans.
Load the data into partitioned BigQuery tables clustered by device_id and serve the dashboard through BI Engine.
Insert each reading into Cloud Spanner tables partitioned on device_id with a secondary index on timestamp for recent queries.

Report Issue

Answer Description

Cloud Bigtable is built for very high write throughput and offers single-digit-millisecond read latency at petabyte scale. A composite row-key such as reverseTimestamp#deviceId keeps the most recent measurements for each sensor adjacent on disk, enabling efficient time-range scans and point lookups that power real-time dashboards. Bigtable clusters scale storage and compute independently and require little administration. BigQuery, Firestore, and Cloud Spanner are all fully managed, but their typical query latencies or cost profiles make them less suitable for this high-ingest, low-latency time-series workload.

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.