GCP Professional Data Engineer Practice Question

Memorystore for Redis is a fully managed, in-memory data store that delivers sub-millisecond latency at high throughput, making it ideal for frequently updated, low-latency caches. A streaming Dataflow job can compute rolling five-minute revenue totals from the transaction stream and write the single aggregated value to Redis on each update. The dashboard then performs simple key lookups against Redis, achieving the required response time with minimal operational work and no per-query BigQuery charges.

BigQuery BI Engine accelerates SQL queries but still relies on BigQuery execution and is unlikely to guarantee sub-100-ms responses for thousands of concurrent users. Cloud SQL and Bigtable can provide millisecond-level latency but are disk-backed, introduce higher query overhead, and are unnecessary for a single rapidly changing aggregate value. Therefore, publishing the aggregate to Memorystore for Redis best meets the latency, cost, and management constraints.