Google Professional Data Engineer Exam Survival Guide

Window: Global | 50–60 scenario Q / 120 min

Must-do topics

• Data ingest: Pub/Sub, Dataflow, Transfer Service, Storage Transfer, Datastream (CDC)
• Storage/Processing: BigQuery (partition/cluster, slots), Cloud Storage, Bigtable, Dataproc
• ML: Vertex AI pipelines, model registry, feature store basics
• Security/Governance: IAM, CMEK, DLP, VPC-SC, fine-grained access
• Reliability/Cost: autoscaling, job windows, reservations, flat-rate vs on-demand

Top traps (avoid)

• Using Dataproc when Dataflow (serverless) fits streaming/batch
• BigQuery scans without partition/cluster → runaway cost
• Bigtable used for analytics instead of OLTP-like wide-column workloads

Time split

• ~2 min/Q; first pass service-fit, second pass edge constraints

Last-48h checklist

• BigQuery storage/compute separation; slot commitments
• Pub/Sub ordering & dead-letter; Dataflow windowing/watermarks
• Security: CMEK vs CSEK; VPC-SC perimeters

Quick facts

• BigQuery ML for in-DB models; BigQuery BI Engine accelerates dashboards
• Datastream = CDC to GCS/BigQuery; Dataplex for governance zones

Speed tactics

• Map requirement → latency/consistency/size → pick service
• Add governance & cost to every design choice

Day-of mini-plan

• Warm-up: 1 ingest + 1 BQ partitioning + 1 security scenario