GCP Professional Data Engineer Practice Question

Your e-commerce company wants to capture click-stream events from mobile clients worldwide and feed them into a streaming analytics pipeline on Google Cloud. Peak traffic is expected to exceed five million events per second. The solution must deliver end-to-end latency under 100 ms, provide a multi-region SLA of at least 99.99 %, automatically scale without advance capacity planning, and require the least operational effort from the data engineering team. Which ingestion service should you implement as the streaming source?

Use Pub/Sub Lite single-zone topics with provisioned throughput partitions.
Write events as JSON files to Cloud Storage buckets for downstream processing.
Deploy a multi-zone Apache Kafka cluster on Google Kubernetes Engine with cluster autoscaling.
Publish events directly to Cloud Pub/Sub multi-region topics.

Report Issue

Answer Description

Cloud Pub/Sub is a fully managed, serverless, global messaging service that automatically scales to millions of messages per second without capacity planning. Multi-region topics provide a 99.99 % availability SLA and built-in cross-zone replication, helping you meet the latency and reliability targets with minimal operational work.

Running Apache Kafka on GKE would meet throughput but leaves cluster provisioning, replication, and upgrades to your team, violating the "least operational effort" requirement.

Pub/Sub Lite lowers cost but is zonal or regional, offers only a 99.95 % SLA, and requires pre-provisioned throughput, so it cannot guarantee the stated latency, availability, or auto-scaling goals.

Writing JSON files to Cloud Storage introduces batch latency that far exceeds the 100 ms threshold and is therefore unsuitable for real-time streaming ingestion.

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.