GCP Professional Data Engineer Practice Question 1: Telemetry Store Selection
Google Cloud Professional Data Engineer · Test 1 · question 1 of 10
A logistics company streams telemetry from 200,000 delivery vehicles. The platform must sustain roughly 1.5 million writes per second, retain 18 months of readings, and serve the fleet-tracking application, which always reads one vehicle's readings over a time range and needs single-digit millisecond latency. No ad hoc SQL analytics run against this store; reporting is handled by a separate warehouse.
Which storage service should you choose for the telemetry store?
Show answer and explanation
Answer:
Bigtable is Google Cloud's wide-column store for high-throughput, low-latency key-value and time-series workloads, and throughput scales in direct proportion to the number of nodes in the cluster. The access pattern in the stem (every read is "one entity, one time range") is exactly what a row key of vehicleID#reverse_timestamp serves: the vehicle ID prefix spreads writes evenly across tablets while keeping one vehicle's readings contiguous, so a range scan returns the newest readings first. Time-series and IoT sensor data are named use cases in the product overview.
Option A does not scale to the write volume. A Cloud SQL instance has a single primary that accepts all writes, and read replicas add read capacity only; millions of writes per second is well past what any single PostgreSQL primary sustains.
Option B misreads Firestore. It is a document database built for application state and mobile or web clients with offline sync, and it has a sustained write limit of about one write per second per document plus per-database throughput characteristics that suit user-facing documents, not raw sensor firehoses.
Option D chooses an analytical warehouse for an operational serving path. The Storage Write API can absorb the ingest volume, but BigQuery is optimized for scanning large amounts of data per query, not for single-digit millisecond point reads of one key, and the stem explicitly places analytics elsewhere.
Reference: Bigtable overview